[MarketDataApp]

Reported from a phone screenshot of MarketData-App/api#463: the timeline is full of marketdata-code-review[bot] referenced this in MarketData-App/code-review-credentials, each showing a lease taken by MarketData… / lease freed by MarketData… commit.

Cause

holder is <owner>/<repo>#<pr>#<run>-<attempt>, and <owner>/<repo>#<pr> is exactly GitHub's cross-repository issue reference syntax. It went into the lease commit message verbatim, so every write to codex/lease.json posted a cross-reference onto the reviewed pull request.

Two events per review, on every review, for the life of the branch.

Measured on api#463 — twelve lease commits, twelve timeline events, matching to the second:

Timeline event	Lease commit
2026-09-22T14:25:26Z	14:25:25 lease taken by MarketData-App/api#463#35740179424-1
2026-09-22T14:26:06Z	14:26:05 lease freed by MarketData-App/api#463#35740179424-1
… 10 more …	…

Its timeline holds 12 referenced events against 2 committed, 2 commented and 3 labeled.

Fix

lease_message spells the same three facts without the #:

lease taken by MarketData-App/api pull 463 run 35743856939-1

The holder is untouched. release compares it to decide whether this run still owns the lease, and that identity is why it carries the run id and attempt — rewriting it would reopen the race #<run_id>-<run_attempt> was added to close. Only the human-facing message changes.

lease_spans now reads both forms and keys a hold on (repo, pr, run) rather than on the message's spelling, so the months of #-joined history already in the store keep parsing. Verified against the live store: 237 holds over 7 days, unchanged.

What this does not do

It cannot retire the events already posted. GitHub builds a cross-reference from the commit and offers no way to delete one. The existing noise on open pull requests stays; no new noise is added.

Tests

11 new, TDD, each watched failing first. 653 passed, ruff clean. One test round-trips lease_message through lease_spans, so a change to either side without the other fails here rather than in next week's report.

[marketdata-code-review]

✅ Ready — Code review

This change removes GitHub cross-repository references from lease commit messages while preserving the stored holder identity. The history parser supports both old and new message formats and normalizes them to the same lease span identity.

Rating patch 5/6 (solid) · proof 6/6 (comprehensive) · overall 5/6

Findings

Nit

• reviewbot/credentials.py:120-129 Move dated incident narrative out of code (style, 1c4d475a)
  The new docstring embeds a dated incident report and measurements that belong in the pull request or commit history. The cited house-style examples do not override the standing rule for newly added comments; retain only the enduring constraint that raw holders create cross-references.

  reviewbot/credentials.py:120-129 — Measured 2026-09-22 on MarketData-App/api#463: twelve lease commits, twelve timeline events, timestamps matching to the second.

Since last review

• Resolved: a1f9e042, 2f9d3e2b
• 0 still open, 1 new

---

Reviewed fb444cd · revision 2 · codex (gpt-5.6-sol) · claude unavailable · rules: default

[MarketDataApp]

a1f9e042 — correct. Fixed.

The holder is now rebuilt once, beside the pairing key, and carried through the open-span map, so both branches append the same string. Old and new formats yield an identical holder, an open hold reads the same as a closed one, and a holder with no pull request stays verification rather than being padded to verification#None#. One test round-trips every shape through lease_message and back.

2f9d3e2b — not taken, same as 8ebd1bc4 and ad392b16 on #19. Dated measurements in docstrings are this repository's house style, not an exception to it: credentials.py:10, backends/claude.py:25, github.py:303, github.py:361, cli.py:42. The date is what makes the claim checkable later, and here the measurement is the evidence that the bug was real — twelve commits, twelve timeline events, matching to the second.

This is the third review running to raise it. If the rule is wanted, it belongs in REVIEW.md where it would apply to the existing instances too; as a per-pull-request nit it just asks new code to differ from the code around it.