docs(disk-hygiene): --execute manual-lane contract, hint-prose trust boundary, doc polish (F7/F9/F10)

[kyle-sexton]

Summary

dh 0.8.1 — the consumer-audit declarative/doc batch (findings F7, F9, F10). Docs-only; no engine or guard code changes.

• F7 — --execute contract for the manual lane (decision recorded, not silently resolved). The flag now explicitly gates every deletion lane: the gated engine lane where supported, the manual handoff elsewhere. This is named in the doc as a deliberate semantic unification — the flag previously read as "offer the gated ENGINE lane" (which can never apply on Windows/macOS), leaving the manual lane's gate ambiguous; the audit observed consumer sessions reading it both ways, including one that proceeded to manual deletion without --execute. The issue's adversarial-review comment asked that the semantic shift be named rather than passed off as clarification — done at the argument definition and in the manual-handoff precondition (lane symmetry, per the issue's recommendation).
• F9 — hint-prose trust boundary. Safety-model trust boundaries now name standing-policy additional_hints[].reason prose as untrusted claims requiring independent evidence: additive-only design means hints can't authorize anything, but the prose reached triage reasoning unlabeled.
• F10(a). Setup SKILL.md + README corrected: preview reports execution-platform-unsupported as a per-candidate blocker, not a top-level status — verified against preview()'s payload shape (blockers live in candidates[].blockers; top-level status is blocked/ready-for-explicit-approval).
• F10(b). README states once that the Recycle-Bin / Trash naming is a model-layer distinction only; the engine treats Windows and macOS identically (execution unsupported).
• F10(c) — DECLINED with evidence. The hub-restructure suggestion rested on an MD013 line-length complaint; the repo's .markdownlint-cli2.jsonc sets "MD013": false ("No hard line-length limit"), so the complaint came from an out-of-repo lint run, and the skill-quality gate passes the hub at its current length.

Scheduling note: #1113 was queued behind draft PR #1124 (file overlap on setup SKILL.md / README / CHANGELOG). That draft has been stale for ~4h with CI green; the actual hunk overlap is disjoint (different sections), so this batch proceeds — whichever merges second rebases trivially (CHANGELOG entry ordering + version bump).

Test plan

• check-changelog-parity.sh --check-bump origin/main → pass (0.8.1 + entry)
• check-changed-skills.sh → 2 skills checked, 0 failed
• Docs-only: no engine tests affected (suite untouched)
• CI green

Related

Closes #1113. Source: handoff-inbox item 20260723-021058-disk-hygiene-0-6-4-consumer-audit, findings F7/F9/F10. F7 decision context: the adversarial-review comment on #1113.

🤖 Generated with Claude Code

[claude]

Claude encountered an error after 1s —— View job

---

I'll analyze this and get back to you.

[claude]

Claude encountered an error after 1s —— View job

---

I'll analyze this and get back to you.

[github-actions]

Warning

Automated security review did not complete — this is an infrastructure failure, not a review verdict.

Treat any Claude comment on this PR (including a placeholder like "I'll analyze this and get back to you") as incomplete, not "no findings."

• Run: https://github.com/melodic-software/claude-code-plugins/actions/runs/30018589672
• Last SDK result: {"subtype":"success","is_error":true,"num_turns":1,"duration_ms":362,"total_cost_usd":0}

Re-running the job, or pushing a new commit, will retry the review.

[github-actions]

Warning

Automated review did not complete — this is an infrastructure failure, not a review verdict.

Treat any Claude comment on this PR (including a placeholder like "I'll analyze this and get back to you") as incomplete, not "no findings."

• Run: https://github.com/melodic-software/claude-code-plugins/actions/runs/30018588957
• Last SDK result: {"subtype":"success","is_error":true,"num_turns":1,"duration_ms":933,"total_cost_usd":0}

Re-running the job, or pushing a new commit, will retry the review.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b9d97f25e1

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[claude]

Claude encountered an error after 1s —— View job

---

I'll analyze this and get back to you.

[claude]

Claude encountered an error after 0s —— View job

---

I'll analyze this and get back to you.