Skip to content

feat(security): add structured credential detectors behind an internal switch - #29369

Merged
mbriones98 merged 1 commit into
mainfrom
mdbrio/redaction-cutover-v2
Nov 11, 2026
Merged

mbriones98 merged 1 commit into
mainfrom
mdbrio/redaction-cutover-v2

Conversation

@mbriones98

@mbriones98 mbriones98 commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Put the credential rule table and its wiring in place behind an internal switch that is off and has no user setting yet, so the table can be finished and checked before anyone can turn it on.

The rule table under security/redaction_rules/ classifies each value by what it can do (Secret, Bearer link, Identifier), as the recorded decision docs/decisions/2026-10-20-redaction-classifies-by-what-a-value-can-do.md sets out. No surface calls it yet: every surface runs the legacy fixed-pattern pass, which redacts an AWS access key ID as if it were a secret.

Why it matters

People reading CloudTrail and service logs through Kiro Crew lose the key ID, the one value that joins a log row to the access key that made the call. A key ID cannot authenticate by itself. This is the first step of the #22552 phase that routes redaction through the table. Shipping it dormant keeps this change small and leaves behaviour exactly as it is today; the user setting arrives in a later change, once the table is ready to be turned on.

Not a goal

  • A user setting (config key, schema entry or dashboard control) for the table. It arrives in the later change that turns the table on.
  • Turning the table on, or updating the tests whose expectations change when it is on.
  • Changing the later passes (base64-encoded credentials, the bare 40-character secret with its link, SRI, ARN and code-text exemptions, the token= parameter pass). They run unchanged either way.
  • The OAuth gate, the exfiltration and decisions gates, and the posture panel wording.
  • Rewriting transcripts already written.

What changed (motivation → approach → change)

Review asked for a structured approach in place of new or wider regex matching. This PR therefore no longer adds regex rows: it adds credential detectors built from structure only, behind the same internal switch, which stays off with no user setting. Main's redaction_rules/ package and its passes are not changed.

Location. The detectors are a new package, src/kiro_crew/security/redaction_structured/, wired behind the switch. Main's unused redaction_rules/ package and its tests are removed, so table.ROWS is the only credential table. Nothing called that package, so the switch-off path still runs exactly the code main runs.

No regex, no randomness scores. The package imports neither re nor regex and has no randomness or character-mix scoring; a source test holds it to both. Rules are data in table.ROWS: names, literal prefixes, alphabets, lengths and checksum ids. The loader refuses a pattern field.

Part What it does
tokens.py, armour.py A token is read from its literal anchor (str.find) over its format's alphabet, with its length, fixed structure and vendor checksum (GitHub CRC32) checked; a JWT/JWE header must decode to JSON naming alg or enc. A 40- or 64-character lowercase hex value is a git object name and is never a Secret; an exact AWS label still wins over it. PEM and PGP private-key blocks are read by their BEGIN/END lines.
known_key.py, identifiers.py The four exact AWS keys (aws_secret_access_key, SecretAccessKey, aws_session_token, SessionToken) in JSON, assignments, CLI flags and XML, with a value-shape check. A URL query is parsed per RFC 3986 and only the values of client_secret, access_token, refresh_token, id_token, code_verifier and password are replaced.
identifier_values.py Marks values that stay visible (AWS key IDs, ARNs, request ids, public OAuth values, data URIs, SSH fingerprints, AWS documentation example keys, placeholders).
stream.py, engine.py, model.py, charsets.py, checks.py, table.py Scanning, hit resolution and streaming. Every unit has a size bound and fails closed past it, so streamed output equals whole-text output at every split.
compat.py The switch, compat.use_rule_table(), set only by code (publish_use_rule_table). Off: the credential plan is pass 1 then passes 2-4, as on main. On: the detectors read the text first, their Secret hits are replaced by the credential tag, the four passes run on that text, and the spans are moved back and merged. A detector error replaces the whole text, logs the error type and counts one metric.
security/redaction.py Splits pass 1 (_pattern_pass) from passes 2-4 so the detectors can run in front of them, and routes the plan by the switch.
security/text_redaction.py StreamRedactor asks the seam where to hold a buffer, and drops a unit decided while still running. Both do nothing while the switch is off.
credential_patterns.py, dashboard/display_verdicts.py The whole-text fail-closed rule (redaction_unavailable) gets its Secret tier; cached display verdicts are keyed on the switch (table=0 while off).

Documented bound. A URL longer than the URL row's 4,096-character bound is not read as a URL by the detectors, so its query values are left to main's credential passes, which run after the detectors and still remove them. A run of mask letters longer than a masked value is decided at once and never held.

docs/system-specs/modules/security.md describes the package, the seam and the switch.

Backwards compatibility

Compatible: the switch is off and no user can turn it on, so redaction output is unchanged; the redaction entry points keep their signatures, plan shape and rule names. No config key is added. Cached display verdicts are recomputed once, because their version now includes the switch value.

The diff removes src/kiro_crew/security/redaction_rules/ and test/test_redaction_rules.py. Its readers on main:

Reader: setup.cfg:package_data -- release -- the rules.yaml package-data line is removed with the package
Reader: scripts/ci_file_shard_pins.txt:test_redaction_rules.py -- release -- the pin for the deleted test file is removed; test_ci_file_shards.py reads the pins
Reader: docs/system-specs/modules/security.md:redaction_rules -- release -- the package's entry is removed

No other module, entry point or app imports it.

Tests

  • test_redaction_structured.py, _git_oauth.py, _identifiers.py, _protect.py, _stream.py cover each detector, the source test (no re/regex import, no randomness or character-mix terms, with self-tests), the six OAuth keys, the AWS label over the git object-name row, and stream-equals-whole-text at every split.
  • Linear work is checked by counting executed lines at two input sizes (assert_linear_work), never by timing, over adversarial inputs including repeated mask runs; a stream fed mask runs in pieces keeps its pending text bounded.
  • test_stream_table_parity.py and test_redaction_compat_property.py turn the switch on through a fixture; one test checks that off runs main's plan.
  • Three guard tests change for the seam: the display battery's import list, the security package line budget (+107, for redaction.py and text_redaction.py; the new subpackage is not counted), and the regex-linearity check, which now reads the split-out pass 1.
  • Off equals main: main's 79 redaction test files pass with the switch off, and the change-scoped backend run (scripts/run_scoped_tests.py --surface backend) is listed below.

Scoped backend run (4,437 related test files, all 8 batches): 298 test ids failed; each also fails on a clean origin/main worktree at the base this was measured on in the same environment (289 in a parallel re-run of the failing files, the other 9 when run one at a time), so none come from this change.

Proof: not provable -- other: a feat change behind an internal switch that is off and has no user setting; its tests run the detectors with the switch on

Manual verification

N/A — unit coverage sufficient: the change is pure text processing behind an internal switch that is off.

Related Issues

Part of #22552

Follow-up: #31895 (one owner setting that turns credential redaction off on every path)

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Known follow-ups

  • Add one owner setting that turns credential redaction off on every path #31895: add one owner setting that turns credential redaction off on every path, bypassing the credential plan and the structured engine, default on.
  • Add the Identifier masking module and the table health check, wire Identifier protection into main's remaining passes, stop the pass-1 branches the detectors cover (keeping main's URL-parameter handling), turn the switch on, and update the test expectations that change.
  • With the switch on, a labelled AWS secret keeps its label in streamed output but not in whole-text output; the secret value is removed either way.

@mbriones98

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 009a7a8
  • What rule changed: No redaction rule changes for any user. This adds a data-driven credential rule table (security/redaction_rules/) and wires it into redaction.py's credential plan and StreamRedactor behind an internal switch, redaction_rules.compat.use_rule_table(). The switch is off and has no config key, CLI flag or API route; only compat.publish_use_rule_table, called from tests, can turn it on. With it off, redaction.py runs its fixed-pattern pass and later passes exactly as before (the function body is split in two, same order and state), and the stream glue does nothing.
  • Before -> after (what was allowed/blocked, now): Unchanged for every surface while the switch is off: the same text is redacted and kept, with the same tags. The only always-on additions are a table=0 component in the display-verdict version key (constant while the switch is off) and an import of the table's stream glue in text_redaction.py, which returns nothing to hold while the switch is off. The display battery is byte-identical to main and the gateway does no table work at start.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): Leak: if the off path differed from main, some credential could go unredacted in chat, cron or subagent output; checked below. Breakage: an import error in the table's stream glue would break streamed replies; it is covered by tests on every run.
  • Evidence checked (which sandbox result proves it): Before the user setting was removed, the full backend suite run the CI way (60 shards) showed no test failing on this branch that passes on main. A reviewer ran 3,852 redaction test inputs through batch redaction and through StreamRedactor with random 1-9 character chunks, switch off, on this code and on main: identical output, warnings and records on every input. test_importing_the_battery_does_not_load_the_rule_table checks, in a clean interpreter, that the display battery loads no table module.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool or command handling changed. The change-scoped local gate ran the related backend tests (5 batches) and frontend tests (21,127 passed); every backend failure it reported also fails on a clean main worktree in the same environment.
    • Backward compatible (old config / old data / old callers still work, how checked): No config key, schema or stored data is added. Existing callers of redact_credentials, redact_credentials_with_records and StreamRedactor keep their signatures. Cached display verdicts gain table=0 in their version key, so they are recomputed once after upgrade.
    • Existing tests (which suites passed on this head): 1,112 passed across test_security_gate_liveness.py, test_facade_mirror_guards.py, test_sections_facade.py, test_config_schema.py, test_display_battery_rules_version.py, test_display_verdicts.py, test_redaction_rules.py, test_stream_table_parity.py, test_redaction_compat_property.py, test_security_regex_linearity.py, test_security_oauth_urls.py; black, isort, flake8 and the file line cap pass.
  • How to undo: Revert this single commit. No config, schema or data migration is involved.

@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Nov 3, 2026
@github-actions

github-actions Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

Sensitive Change Review

Sensitive files changed:

  • src/kiro_crew/security/redaction.py

  • src/kiro_crew/security/redaction_rules/__init__.py

  • src/kiro_crew/security/redaction_rules/detectors.py

  • src/kiro_crew/security/redaction_rules/model.py

  • src/kiro_crew/security/redaction_rules/redact.py

  • src/kiro_crew/security/redaction_rules/rules.py

  • src/kiro_crew/security/redaction_rules/rules.yaml

  • src/kiro_crew/security/redaction_structured/__init__.py

  • src/kiro_crew/security/redaction_structured/armour.py

  • src/kiro_crew/security/redaction_structured/charsets.py

  • src/kiro_crew/security/redaction_structured/checks.py

  • src/kiro_crew/security/redaction_structured/compat.py

  • src/kiro_crew/security/redaction_structured/engine.py

  • src/kiro_crew/security/redaction_structured/identifier_values.py

  • src/kiro_crew/security/redaction_structured/identifiers.py

  • src/kiro_crew/security/redaction_structured/known_key.py

  • src/kiro_crew/security/redaction_structured/model.py

  • src/kiro_crew/security/redaction_structured/stream.py

  • src/kiro_crew/security/redaction_structured/table.py

  • src/kiro_crew/security/redaction_structured/tokens.py

  • src/kiro_crew/security/text_redaction.py

  • @mbriones98 posted a complete reasoning template for b828241.

@github-actions

github-actions Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ⏭️ skipped

Revision b8282415a6ef65f10f4c9385c45ca2888c0af7c0 touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Does not block merge.

@github-actions

github-actions Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of b8282415a6ef65f10f4c9385c45ca2888c0af7c0 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: with the switch off, the redaction path behaves as it does on main, and removing redaction_rules/ leaves no reader in this tree.

Notes

  • About 3,500 lines of detectors now ship with no user setting and no production code path that turns them on. Until the change that enables them lands, nothing outside their own tests keeps them in step with main's passes, so they can drift out of date. Keep that follow-up PR close behind this one.
  • Turning the switch on keeps AWS key IDs visible, which deliberately weakens today's redaction. The enabling PR should say this under ## Backwards compatibility and cite the 2026-10-20 decision for it.

[DESIGN-REVIEWED] b828241

Description read

[DESCRIPTION-READ] c3bc7f219032ea097fd2e08f69780507dce4d9ddc6da9c7c19b7bba05bb8674e

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of b8282415a6ef65f10f4c9385c45ca2888c0af7c0 — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

First-Principles-Verdict: PASS

Nothing blocks, because the code is dormant and easy to remove. But even with the switch on, this wiring cannot deliver the harm the Goal names, since pass 1 still redacts key IDs.

Notes

  • Item 3 — symptom-level: switched on, _combine still runs redaction._main_plan (pass 1 included). Pass 1's (re.compile(AWS_KEY_ID), "aws_access_key_id") in _PASS1_RULES then redacts the AKIA/ASIA IDs that the identifier_values.py rows mark as visible. That pass-1 branch is what causes "people lose the key ID". Subtraction: remove the AWS_KEY_ID branch from the pass-1 path while the switch is on.
  • Item 2 — duplicate of security/redaction.py:_PASS1_RULES: _OLD_RULE maps 18 table rows onto pass-1 rule names, and pass 1 detects the same formats again in the same plan. The decision 2026-10-20 says "One rule table holds every tier and pattern… No surface keeps its own copy." Subtraction: on the switched-on path, drop the pass-1 branches those 18 rows already cover.
  • Item 1 — zero consumers: publish_use_rule_table has 0 callers in src/ outside compat.py (grep publish_use_rule_table). Only tests call it, so about 3,050 added lines run only under test until a later PR adds the setting.

What this change ships

Inventory (6 items) — 3 justified

Intent: stage a credential classifier that knows a value's capability, behind a switch that is off, as the first phase of #22552. This is an ADDITION.

  1. Internal switch compat.use_rule_table(), off, settable only from code — zero consumers, only tests set it
  2. When on, structured detectors run ahead of the four legacy passes — duplicate of _PASS1_RULES for 18 formats
  3. Identifier tier (AWS key IDs, ARNs, request IDs) kept visible — symptom-level, pass 1 still redacts key IDs
  4. A detector failure replaces the whole text and counts a metric — justified
  5. Unused redaction_rules/ package, its YAML and its tests removed — justified
  6. Cached display verdicts keyed on the switch, so they are recomputed once — justified

[FIRST-PRINCIPLES-REVIEWED] b828241

Description read

[DESCRIPTION-READ] e8d457f703b0d61a2c1c4a4d4dc4a77ae499ea3b6babe8c73e307bd6071b15d8

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed b8282415a6ef65f10f4c9385c45ca2888c0af7c0 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] b828241

Verdict parsed from the review's SHA-scoped output markers for commit b8282415a6ef65f10f4c9385c45ca2888c0af7c0.

False positive or not applicable? A repository writer can comment:
/ai-review override fable b8282415a6ef65f10f4c9385c45ca2888c0af7c0 false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of b8282415a6ef65f10f4c9385c45ca2888c0af7c0 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] b828241

False positive or not applicable? A repository writer can comment:
/ai-review override gpt b8282415a6ef65f10f4c9385c45ca2888c0af7c0 false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

Security Scope Review -- ✅ PASS

Which legitimate operations does b8282415a6ef65f10f4c9385c45ca2888c0af7c0 newly refuse? Updated in place on each push. A script-confirmed regression blocks; the model proposes candidates and never decides a refusal.

Verdict basis: adjudicated, zero confirmed regressions.

Security scope review -- adjudicated candidates

  • macos-latest (platform posix): 13 classified, 0 newly refused, 12 still allowed.
  • ubuntu-latest (platform posix): 13 classified, 0 newly refused, 12 still allowed.
  • windows-latest (platform windows): 8 classified, 0 newly refused, 8 still allowed.

No candidate flipped from allowed to refused.

Scope-Verdict: PASS

Reason: This PR deletes the unused redaction_rules/ table and adds a structured detector package, redaction_structured/. The package only runs when the internal compat.use_rule_table() switch is on, and that switch defaults to off with no user setting, so the shell deny composite's behaviour should be unchanged. The detectors' behaviour with the switch on is a redaction guard the classifier cannot judge; it is named below under UNADJUDICATED.

TIGHTENING: structured credential detectors placed in front of the redaction credential plan -- credentials that the fixed patterns miss leaking through redacted output. The hunk is in redaction.py _credential_redaction_plan: if compat.use_rule_table(): return compat.credential_plan_via_table(text, defer_token_param, judged_sri_values, token_param_prefixes). Its fail-closed path replaces the whole text with REDACTION_UNAVAILABLE_TAG = "[REDACTED: redaction unavailable]".

ADJUDICATED: 17 candidates, 0 confirmed newly-refused

PLATFORM COVERAGE: Rows are written for posix, windows and any. The diff assumes no path separator, interpreter or environment-variable spelling, so the platform rows only probe the exfil tier, which calls redact_credentials.

  • posix: a token taken from $GITHUB_TOKEN in a curl Authorization header, AWS identity and profile lookups, an SSH fingerprint check, a certificate read, an OAuth authorize URL with only public query values, a connection-string host read, and a pytest run.
  • windows: the same token header as %GITHUB_TOKEN% with curl.exe, ssh-keygen.exe with %USERPROFILE%\ paths, the AWS lookup, and .venv\Scripts\python.exe with \ test paths.
  • any: 40-hex and 64-hex git object names (the new package says these are never a secret), a remote-URL read, and a grep for redaction tags.

No platform the diff assumes is left without a row.

UNADJUDICATED:

  • The redaction plan with the switch on. Call path: compat.credential_plan_via_table → engine/table.ROWS detectors. It is reached through redaction._credential_redaction_plan, which runs redact_credentials, redact_credentials_with_records, pass_compose and StreamRedactor (through compat.stream_read). That covers chat, dashboard, cron output and exfil._credential_clean / _blocked_link_safe_domain for blocked-link records. Rows can't express the legitimate text I would have submitted, because the classifier only judges whether a command is refused, not what redaction does to output:
    • a 40-hex or 64-hex commit hash in git log output;
    • a JWT-looking string that is three dot-joined base64url segments but whose header does not decode to JSON with alg/enc;
    • a PEM certificate or public-key block, which is not a PRIVATE KEY block;
    • an OAuth redirect whose query carries code, state or client_id, but none of the six secret keys;
    • an AWS key ID shown on its own, without its secret;
    • an SNS topic ARN whose account ID is followed by a long CDK-generated topic name;
    • documentation example keys;
    • a stream chunk cut in the middle of a BEGIN … PRIVATE KEY line.
  • The fail-closed path. With the switch on, any detector error or table-load failure replaces the whole text with [REDACTION_UNAVAILABLE_TAG] and makes every later plan in the process fail closed (_LOAD_FAILED). The candidate input is a long but harmless log or transcript that goes past a unit's size bound, which "fails closed past it". Whether that replaces the whole text, rather than just the unit, can't be checked here and should be tested before the switch is turned on.

[SCOPE-REVIEWED] b828241

@mbriones98

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 37ea136
  • What rule changed: No redaction rule changes for any user. This adds a data-driven credential rule table (security/redaction_rules/) and wires it into redaction.py's credential plan and StreamRedactor behind an internal switch, redaction_rules.compat.use_rule_table(). The switch is off and has no config key, CLI flag or API route; only compat.publish_use_rule_table, called from tests, can turn it on. With it off, redaction.py runs its fixed-pattern pass and later passes exactly as before (the function body is split in two, same order and state), and the stream glue does nothing. The second commit changes no behaviour: it writes three PEM test fixtures as split string literals, as other tests do, and marks two log calls in compat.py that log only an exception type with the rule's nosemgrep reason.
  • Before -> after (what was allowed/blocked, now): Unchanged for every surface while the switch is off: the same text is redacted and kept, with the same tags. The only always-on additions are a table=0 component in the display-verdict version key (constant while the switch is off) and an import of the table's stream glue in text_redaction.py, which returns nothing to hold while the switch is off. The display battery is byte-identical to main and the gateway does no table work at start.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): Leak: if the off path differed from main, some credential could go unredacted in chat, cron or subagent output; checked below. Breakage: an import error in the table's stream glue would break streamed replies; it is covered by tests on every run.
  • Evidence checked (which sandbox result proves it): Before the user setting was removed, the full backend suite run the CI way (60 shards) showed no test failing on this branch that passes on main. A reviewer ran 3,852 redaction test inputs through batch redaction and through StreamRedactor with random 1-9 character chunks, switch off, on this code and on main: identical output, warnings and records on every input. test_importing_the_battery_does_not_load_the_rule_table checks, in a clean interpreter, that the display battery loads no table module.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool or command handling changed. The change-scoped local gate ran the related backend tests (5 batches) and frontend tests (21,127 passed); every backend failure it reported also fails on a clean main worktree in the same environment.
    • Backward compatible (old config / old data / old callers still work, how checked): No config key, schema or stored data is added. Existing callers of redact_credentials, redact_credentials_with_records and StreamRedactor keep their signatures. Cached display verdicts gain table=0 in their version key, so they are recomputed once after upgrade.
    • Existing tests (which suites passed on this head): 1,112 passed across test_security_gate_liveness.py, test_facade_mirror_guards.py, test_sections_facade.py, test_config_schema.py, test_display_battery_rules_version.py, test_display_verdicts.py, test_redaction_rules.py, test_stream_table_parity.py, test_redaction_compat_property.py, test_security_regex_linearity.py, test_security_oauth_urls.py; black, isort, flake8 and the file line cap pass.
  • How to undo: Revert both commits. No config, schema or data migration is involved.

@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed readiness: checking Automated validation is still running labels Nov 4, 2026

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks good to me at 37ea136, and I would approve it, but it now conflicts with main. Please rebase and I will approve the new head.

What I checked, since this touches the credential redactor every surface uses:

  • The switch is off and nothing outside code turns it on. compat.use_rule_table() routes the credential plan's first pass to the table; with it off, _credential_redaction_plan runs _pattern_pass and then _later_passes, which is the old body split in two with no change to either half.
  • The stream redactor is unchanged while the switch is off. StreamRedactor now calls redaction_rules.stream on every chunk, so I read each call. read_buffer returns the empty result when table_active() is None, and keep_pairs / floor_past_pairs then return the cut unchanged, so table.open and table.strong are empty and the discard and hold logic takes the old branches. redact_window adds no prefix when the table is off. discard_run_res and discard_carry are not gated, but every kind they add is named table-*, so they cannot replace the four legacy discard patterns or change a legacy carry. line_context only tracks the current line.
  • Small side effects: rules_version() now hashes table=0, so the display verdict version changes once on upgrade; the new redaction_unavailable tier and its empty-value record are reached only through the table path.

All 100 checks are green, including the Security Scope Review, the redaction property and parity tests, and the regex linearity check.

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Requesting changes at 37ea136. This replaces my earlier comment that said I would approve after a rebase.

The maintainers have adopted a rule that we do not approve a change whose solution relies on adding or widening regex matching to redact secrets, scan output, or build rules; we ask for a structured approach instead. The credential rule table is built that way: its detection is new regex throughout, with about 22 new pattern: entries in security/redaction_rules/rules.yaml and new compiled patterns in redaction_rules/detectors.py, redaction_rules/stream.py and redaction_rules/compat.py. Shipping it behind a switch that is off does not change that, since the table is the design the later change turns on.

Please either move detection to a structured design, for example exact key matching on known field names (the AccessKeyId / SecretAccessKey style labels the table already keys on), a typed scanner that tokenises the text and classifies whole tokens by their known formats, or parsing the structured inputs (JSON, headers, query strings) and judging values by key; or raise the approach with the maintainers first, so the rule table's design can be agreed before more of it lands.

What I checked before and still holds: with the switch off, the credential plan and the stream redactor take the same path as main, and no config, start step or dashboard path turns the switch on. The PR also conflicts with main now.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: passed Eligible automated validation passed for the current revision labels Nov 4, 2026
@mbriones98
mbriones98 marked this pull request as draft November 4, 2026 17:07
@mbriones98

Copy link
Copy Markdown
Contributor Author

@bolichen97 thanks. Before I push more code, here is the structured design I plan to put in this PR and the follow-up. Your rule, as you stated it:

we do not approve a change whose solution relies on adding or widening regex matching to redact secrets, scan output, or build rules; we ask for a structured approach instead.

Detection, with no regex and no entropy

Nothing in the new package imports re or regex, and a source test enforces that. It also adds no randomness scores or character-mix checks. Rules are data: names, prefixes, alphabets, lengths and checksum ids. The loader rejects a pattern field.

  • Typed token scanner. Each format has a literal anchor (ghp_, xoxb-, sk-ant-, glpat-, eyJ, AKIA, ...) found with str.find. The scanner reads the whole token over the format's alphabet and checks the prefix, the alphabet, the exact or published length, any fixed structure, and the vendor checksum where one exists (GitHub CRC32). JWT and JWE are split on ., and the header must decode to JSON with alg or enc. A 40- or 64-character lowercase hex token is a git object name and is never redacted.

  • Private-key blocks. Found by their fixed BEGIN/END armour lines.

  • Exact keys in parsed input. Only aws_secret_access_key, SecretAccessKey, aws_session_token and SessionToken are known keys. They are read from JSON, assignments, CLI flags and XML, and the value must have the AWS secret or session token shape. URL query strings are parsed per RFC 3986. Only the values of client_secret, access_token, refresh_token, id_token, code_verifier and password are redacted. code, state, token, client_id and the rest stay.

  • What is gone. There is no generic label rule (password=, token=, Authorization:), no cookie rule, no Telegram, Discord or PyPI rows, and no open-length fallbacks.

  • Identifier protection. The same scanner marks values that must stay visible, using structure only:

    • AWS key IDs, ARNs, x-amz-id-2 and request ids;
    • public OAuth values, data URIs and SSH fingerprints;
    • AWS documentation example keys and placeholders.

    Main's remaining credential passes do not touch these spans. Protection only takes redaction away.

  • Streaming. Every unit has a size bound. Past the bound, it fails closed by replacing the value and discarding to the unit's end. No unit reads past its own end, so streamed output equals whole-text output by construction.

Measured against main

These are local corpora rebuilt from the earlier experiments: a 9,000-case labelled synthetic set (the first four rows), 611 benign prose records, and a reconstruction of your 52-case simulation. 31 of those 52 cases are my guesses, since the review gave totals, so please share the originals if you can.

main proposed
Secret recall .748 .811
Precision .656 .945
Values that had to stay, removed (break rate) .319 .038
Benign prose, break rate .320 .107
52 cases: benign wrongly redacted / secrets missed 15/35, 3/17 1/35, 1/17
Streamed output differs from whole-text output (322 records, every split), scanner alone 19 records, 14 showing a secret 0
  • Speed. The scanner alone runs at about 2.5 times main's speed on the synthetic set, and is faster than main on short records (19 vs 38 µs) and on benign prose (14 vs 15 µs). Until the covered branches of main's pass 1 are removed, the composed stack runs both and is slower than main (about 1.9 vs 2.5 MB/s).
  • Streaming in the composed stack. Main's passes that keep running keep their own stream holds. For example, main's URL-password branch differs between streamed and whole-text output (5 of the 19 records), and that stays until the branch is decided (question 2).
  • Main's own tests. Measured on an earlier commit of the scanner: with it wired in place of the pass-1 branches it covers, 359 of 3,229 tests in main's 74 redaction test files change outcome. All but one are intended. Most are AWS key IDs staying visible; the next largest groups are the AWS value-shape check and GitHub tokens with a bad checksum. The one bug is fixed.

Not changed

The exfiltration URL check is not changed, so Crew-built sign-in links are still removed whole, as on main. Making that check remove only the parameter it objects to needs its own proposal, and I'd raise it separately.

Plan for the two PRs

I'd rebuild both PRs in place:

  • This PR: the engine seam plus the structured detectors, with the switch still off.
  • The follow-up: identifier protection, turning the switch on, and the test expectation updates. Every difference from main is listed for review.

Questions for the maintainers

  1. Where is the rule recorded, so the code and docs can cite it?
  2. Main's own passes are regex too: pass 1's branches the table doesn't cover, pass 2 (base64-decoded credentials), pass 3 (entropy) and pass 4 (?token=). I plan to leave them running unchanged and never widen them. Is that acceptable, or should any be retired? In my measurements, turning off pass 3 costs about 8 points of recall, and turning off passes 2 and 4 costs about 3 more.
  3. Is this PR the right place for this, or would you rather have it as an RFC in docs/request-for-change/ first? The full write-up is ready to add.

@mbriones98 mbriones98 changed the title feat(security): add the credential rule table behind an internal switch feat(security): add structured credential detectors behind an internal switch Nov 5, 2026
@mbriones98
mbriones98 force-pushed the mdbrio/redaction-cutover-v2 branch from 37ea136 to c63020d Compare November 5, 2026 04:25
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: checking Automated validation is still running labels Nov 5, 2026
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: checking Automated validation is still running labels Nov 10, 2026
@mbriones98

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 72779a5
  • What rule changed: No redaction rule changes for any user. This head is the same feature commit (61660b91a9) squashed onto current main (b7b7b6ccb3) as one commit, folding the earlier merge of main (aae28f054e). It adds the structured credential detectors (security/redaction_structured/), wired into redaction.py's credential plan and StreamRedactor behind the internal switch redaction_structured.compat.use_rule_table(), which is off and has no config key, CLI flag or API route. The squash had one conflict, in test/test_security_gate_liveness.py (the _PACKAGE_LINE_BUDGET hunk): main's two new budget notes and this PR's note are all kept, and the budget is set to the re-measured package size, 35,346 lines.
  • Before -> after (what was allowed/blocked, now): Unchanged while the switch is off. The squashed commit touches the same 37 files as the previous head. Its diff against main, with the liveness file excluded, has the same git patch-id --stable as the previous head's diff against its merge base (fa2b339a946d). The only other difference is the budget hunk above.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): Leak: if the conflict resolution had dropped one of main's redaction changes, a value main redacts could pass through. The only conflicted file is a test constant, and redaction.py merged without a conflict. Breakage: a budget set too low fails test_the_package_stays_within_its_line_budget; it passes at 35,346.
  • Evidence checked (which sandbox result proves it): Run locally on the squashed tree: 4,817 passed and 6 skipped across 101 redaction, credential and security test files, including test_security_gate_liveness.py (10 passed), test_redaction_compat_property.py, test_stream_table_parity.py, test_redaction_structured*.py and test_stream_redactor_aws_keyed.py. On this head in CI, Security Scope Review passed on Linux, macOS and Windows. The backend test reds on this head are 12 Slack code-channel tests and test_session_markdown.py::...[long-s]; main's own CI run on ed80c1d0f2 fails the same 13 tests, and this PR changes no Slack or session-Markdown file.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool or command handling changed.
    • Backward compatible (old config / old data / old callers still work, how checked): No config, schema or stored data changes; the file list and patch identity match the previous head.
    • Existing tests (which suites passed on this head): the suites listed above; black, isort, flake8 and mypy pass on the changed files, the comment-history gate passes against main, and scripts/docs-lint.sh exits 0.
  • How to undo: Revert the single commit to remove the detectors. No config, schema or data migration is involved.

Written by the Crew Tasks Bot on behalf of mbriones98.

@mbriones98
mbriones98 force-pushed the mdbrio/redaction-cutover-v2 branch from 72779a5 to 12d7b05 Compare November 10, 2026 21:44
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Nov 10, 2026
@mbriones98

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 12d7b05
  • What rule changed: No redaction rule changes for any user. This head is the same feature commit squashed onto current main (fd016efd02) as one commit, folding the earlier merges of main, plus one fix from the Opus 5.5 Review on 72779a5600. It adds the structured credential detectors (security/redaction_structured/), wired into redaction.py's credential plan and StreamRedactor behind the internal switch redaction_structured.compat.use_rule_table(), which is off and has no config key, CLI flag or API route.
    • Opus fix: with the switch on, StreamRedactor.feed lowered the holdback floor to the detectors' cut on every feed, so a long run of credential-class characters with no anchor was held until flush(). compat.StreamRead now carries undecided, set only when the cut stops at a unit the detectors have not decided yet (an open unit, or a hit the cut would split). text_redaction.py lowers the floor only then, and only to the detectors' raw cut, so such a run keeps the 512-character bound.
    • Conflict resolution: the only file that needed a change after merging main was test/test_security_gate_liveness.py (_PACKAGE_LINE_BUDGET). Main's two raise notes are kept, and the budget is the re-measured package size, 35,348 lines.
  • Before -> after (what was allowed/blocked, now): Unchanged while the switch is off: stream_read returns None and the new branch is not reached. Outside text_redaction.py, redaction_structured/compat.py, test/test_stream_table_parity.py and test/test_security_gate_liveness.py, this commit's diff against main has the same git patch-id --stable as the previous head's diff against its base, over the same 37 files. With the switch on, 40 feeds of 256 letters now emit 9,728 characters and hold 512, the same as with the switch off; before the fix they emitted 0 and held 10,240.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): Leak (switch on only): if undecided were false for a unit the detectors are still reading, the floor could commit part of that unit. It is set wherever the cut was moved for an open unit or a hit the cut would split, the same two cases that moved the cut before. Breakage: none while the switch is off.
  • Evidence checked (which sandbox result proves it): Run locally on this commit: the new test_an_anchorless_run_keeps_the_holdback_bound failed before the fix (assert 768 <= 512) and passes after it; 6,113 passed and 6 skipped across 140 redaction, credential, security, stream and Slack code-channel test files, including test_security_gate_liveness.py (10 passed), test_stream_table_parity.py, test_redaction_structured_stream.py, test_redaction_compat_property.py and test_stream_redactor_aws_keyed.py.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool or command handling changed.
    • Backward compatible (old config / old data / old callers still work, how checked): No config, schema or stored data changes. StreamRead.undecided defaults to False and the class has no caller outside compat and text_redaction.
    • Existing tests (which suites passed on this head): the suites listed above; black, isort, flake8 and mypy pass on the changed files, the comment-history gate passes against main, and scripts/docs-lint.sh exits 0.
  • How to undo: Revert the single commit to remove the detectors and the fix. No config, schema or data migration is involved.

Written by the Crew Tasks Bot on behalf of mbriones98.

cruisercohen
cruisercohen previously approved these changes Nov 10, 2026
@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Nov 10, 2026
@mbriones98

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: ab9384a
  • What rule changed: No redaction rule changes for any user. This head is a merge of main (4582dce272) into the reviewed head 12d7b05908; its first parent is 12d7b05908 and its second is 4582dce272, which is on main. The feature commit is unchanged: the structured credential detectors (security/redaction_structured/), wired into redaction.py's credential plan and StreamRedactor behind the internal switch redaction_structured.compat.use_rule_table(), which is off and has no config key, CLI flag or API route, plus the stream holdback fix (StreamRead.undecided) from the Opus 5.5 Review on 72779a5600.
  • Before -> after (what was allowed/blocked, now): Unchanged from 12d7b05908. The merge was clean: git merge-tree of the two parents reports no conflict and gives exactly this head's tree. The PR's own diff (4582dce272..ab9384ad80) touches the same 37 files as 12d7b05908's diff against its base and has the same git patch-id --stable (6a3c9ec52f53). The 37 files are byte-identical between 12d7b05908 and this head, so main changed none of them.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): Leak: if the merge had changed a redaction file, a value could be judged differently from the reviewed head. None of the PR's files differ between the two heads, and the merge brought in only main's own commits.
  • Evidence checked (which sandbox result proves it): git log of the merge's parents, git merge-tree --write-tree of those parents compared with this head's tree, git diff --name-only and git patch-id --stable of the PR diff on both heads, and git diff of the 37 PR files between the heads (empty). The tests and lint reported on 12d7b05908 (6,113 passed and 6 skipped across 140 redaction, credential, security, stream and Slack code-channel test files; black, isort, flake8, mypy, the comment-history gate and docs-lint) were run on that head and not re-run on this merge.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool or command handling changed.
    • Backward compatible (old config / old data / old callers still work, how checked): No config, schema or stored data changes; the PR's files are identical to the reviewed head.
    • Existing tests (which suites passed on this head): not run locally on this head; see the evidence line. CI on this head has not produced test verdicts yet because its lanes failed on GitHub's API rate limit.
  • How to undo: Revert the PR's changes as one squash-merge; the merge of main itself changes no PR file. No config, schema or data migration is involved.

Written by the Crew Tasks Bot on behalf of mbriones98.

@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Nov 11, 2026
@mbriones98
mbriones98 force-pushed the mdbrio/redaction-cutover-v2 branch from aa31d97 to 4a1932d Compare November 11, 2026 01:10
@mbriones98

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 4a1932d
  • What rule changed: No redaction rule changes for any user. This head is the same feature commit squashed back to one commit on current main (c1d1a5e32c), replacing the three-commit branch (12d7b05908 plus the two "Merge branch main" commits ab9384ad80 and aa31d97169). It carries the structured credential detectors (security/redaction_structured/), wired into redaction.py's credential plan and StreamRedactor behind the internal switch redaction_structured.compat.use_rule_table(), which is off and has no config key, CLI flag or API route, and the stream holdback fix from the Opus 5.5 Review on 72779a5600: compat.StreamRead.undecided is set only when the cut stops at a unit the detectors have not decided yet, and text_redaction.py lowers the holdback floor only then, with the regression test test_an_anchorless_run_keeps_the_holdback_bound in test/test_stream_table_parity.py.
    • Conflict resolution: merging main raised no conflict. test/test_security_gate_liveness.py keeps main's raise notes and the budget stays at the re-measured package size, 35,348 lines.
  • Before -> after (what was allowed/blocked, now): Unchanged from 12d7b05908. This commit's diff against main touches only the PR's 37 files, has the same git patch-id --stable as 12d7b05908's diff (6a3c9ec52f53), and the 37 files are byte-identical to 12d7b05908 (git diff --stat between them is empty).
  • Worst case if wrong (breaks commands / leaks secret / locks user out): Leak: if the squash had dropped or changed a PR file, a value could be judged differently from the reviewed head. The file list, patch identity and file contents above match 12d7b05908.
  • Evidence checked (which sandbox result proves it): Run locally on this squashed tree: 6,113 passed and 6 skipped across 140 redaction, credential, security, stream and Slack code-channel test files, including test_security_gate_liveness.py (10 passed), test_stream_table_parity.py, test_redaction_structured_stream.py and test_redaction_compat_property.py. After re-placing the commit on the newer main c1d1a5e32c (git range-diff shows it unchanged; no newer main commit touches the security package), those four files passed again (399 passed). push_guard.py --require-single-on-base reports one commit on main.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool or command handling changed.
    • Backward compatible (old config / old data / old callers still work, how checked): No config, schema or stored data changes; the PR's files match the reviewed head.
    • Existing tests (which suites passed on this head): the suites listed above; black, isort, flake8 and mypy pass on the changed files, the comment-history gate passes against main, and scripts/docs-lint.sh exits 0.
  • How to undo: Revert the single commit to remove the detectors and the fix. No config, schema or data migration is involved.

Written by the Crew Tasks Bot on behalf of mbriones98.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed readiness: checking Automated validation is still running labels Nov 11, 2026
…l switch

Add `security/redaction_structured/`, credential detectors built from
structure only: no regular expressions and no randomness or
character-mix scores. A source test holds the package to both. Rules
are data in `table.ROWS` (names, literal prefixes, alphabets, lengths,
checksum ids) and the loader refuses a `pattern` field.

- `tokens.py` reads a token from its literal anchor over its format's
  alphabet and checks its length, fixed structure and vendor checksum
  (GitHub CRC32). A JWT or JWE header must decode to JSON naming `alg`
  or `enc`. A 40- or 64-character lowercase hex value is a git object
  name and is never a Secret.
- `armour.py` reads PEM and PGP private-key blocks by their BEGIN/END
  lines.
- `known_key.py` reads the four AWS secret key names in JSON,
  assignments, CLI flags and XML, with a value-shape check.
- `identifiers.py` parses a URL query per RFC 3986 and replaces only
  the values of six OAuth keys. It and `identifier_values.py` mark
  values that stay visible.
- Every unit has a size bound and fails closed past it, so `stream.py`
  gives the whole-text result for any split. A run of mask letters
  longer than a masked value is decided at once and never held.
- A URL longer than the URL row's 4096-character bound is not read as
  a URL, so its query values are left to the credential passes that
  run after the detectors.

`compat.py` is the seam the credential plan reads. Its switch,
`compat.use_rule_table()`, is off and has no user setting. Off, the plan
is pass 1 then passes 2-4, as before. On, the detectors read the text
first, their Secret hits are replaced by the credential tag, the four
passes run on that text, and the spans are moved back and merged. A
detector error replaces the whole text, logs the error type and counts
one metric. `StreamRedactor` asks the seam where to hold a buffer. The
display verdict version keys the switch value.

Remove `security/redaction_rules/` and its tests, so `table.ROWS` is
the only credential table. Nothing called it: no seam, gateway step or
user setting read it.

With the switch on, only a unit the detectors are still deciding (an
open unit, or a hit the cut would split) lowers the stream holdback
floor, and only to the detectors' own cut. `StreamRead.undecided` says
which. A long run with no anchor now keeps the 512-character bound
instead of being held until flush.

This commit folds the earlier merges of main into one commit on current
main. The package line budget in test_security_gate_liveness.py keeps
main's raise notes and is the re-measured size, 35,533 lines.
@mbriones98

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: b828241
  • What rule changed: No redaction rule changes for any user. This head is the same feature commit as 4a1932d174, squashed again as one commit onto current main (a8308a1ae5). It carries the structured credential detectors (security/redaction_structured/), wired into redaction.py's credential plan and StreamRedactor behind the internal switch redaction_structured.compat.use_rule_table(), which is off and has no config key, CLI flag or API route, and the stream holdback fix from the Opus 5.5 Review on 72779a5600: compat.StreamRead.undecided is set only when the cut stops at a unit the detectors have not decided yet, and text_redaction.py lowers the holdback floor only then, with the regression test test_an_anchorless_run_keeps_the_holdback_bound in test/test_stream_table_parity.py.
    • Conflict resolution: the only conflict was the _PACKAGE_LINE_BUDGET hunk in test/test_security_gate_liveness.py. Main's three newer raise notes (to 35,411) are kept, followed by this PR's note, and the budget is the re-measured package size, 35,533 lines (122 over main). docs/system-specs/modules/security.md merged without a conflict.
  • Before -> after (what was allowed/blocked, now): Unchanged from 4a1932d174. Excluding the liveness budget hunk, this commit's diff against main has the same git patch-id --stable as 4a1932d174's diff (5eb38911a682) and touches the same 37 files. The only PR file whose full content differs from 4a1932d174 is docs/system-specs/modules/security.md, because main changed other parts of it.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): Leak: if the squash had dropped or changed a PR hunk, a value could be judged differently from the reviewed head. The patch identity above matches. Breakage: a budget set too low fails test_the_package_stays_within_its_line_budget; it passes at 35,533.
  • Evidence checked (which sandbox result proves it): Run locally on the squashed tree on main 5544d8e602: 6,113 passed and 6 skipped across 140 redaction, credential, security, stream and Slack code-channel test files, including test_security_gate_liveness.py. After re-placing the commit on the newer main a8308a1ae5 (git range-diff shows it unchanged; no newer main commit touches the security package or the redaction tests), test_security_gate_liveness.py, test_stream_table_parity.py, test_redaction_structured_stream.py and test_redaction_compat_property.py passed again (399 passed). push_guard.py --require-single-on-base reports one commit on main.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool or command handling changed.
    • Backward compatible (old config / old data / old callers still work, how checked): No config, schema or stored data changes.
    • Existing tests (which suites passed on this head): the suites listed above; black, isort, flake8 and mypy pass on the changed files, the comment-history gate passes against main, and scripts/docs-lint.sh exits 0.
  • How to undo: Revert the single commit to remove the detectors and the fix. No config, schema or data migration is involved.

Written by the Crew Tasks Bot on behalf of mbriones98.

@github-actions github-actions Bot removed the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Nov 11, 2026
@mbriones98
mbriones98 disabled auto-merge November 11, 2026 18:36
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

posix-only-approved Cross-Platform Portability findings reviewed and accepted as intentionally POSIX-only readiness: checking Automated validation is still running

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants