[djbclark]

Follow-up to PR #5 (base fix/secrets-collect-yes, PR 5 still open): the four low findings F1 to F4 deferred by its round-4 APPROVE.

1. F1: the SKILL's follow-up printf 'cleared' | clip no longer runs after a run that cleared (it erased the consumer name in the sentinel). It runs only on exit 3, NOT cleared in stdout or stderr, or no report at all; a run that never read the clipboard is fixed and re-run, and the printf is only for giving up on the name.
2. F2: in-process test of the production lock path (pwd.getpwuid faked, no env override, TMPDIR elsewhere): ~/Library/Caches/secrets-collect.lock, dir mode 0700.
3. F3: a run that stored (or dry-ran) but could not clear the clipboard exits 3, with a last line saying so. Other failures keep 1, 2, 130.
4. F4: the sentinel carries date, time and zone (2026-10-26T01:40:59-0400) and the other-name error adds its age. Round 5 BLOCKed a first version that turned a stale sentinel into "no re-collection needed" (a wrong-name store was reachable with latency over 60 s); the age is now information only and the other-name text is always conditional.

Adversary rounds (read-only, omp Opus 5.5 high; reports in ~/.local/state/bigteam/secrets-collect-yes/):

1. Round 5 on 37c8044: BLOCK (reviewed head 37c80446db56b20ee9105d3840dedfadfe8f03d5, delta f6e6208..37c8044): the F4 fix makes the dangerous direction possible. Repaired in 65b51e8.
2. Round 6 on 65b51e8, quoted: APPROVE (reviewed head 65b51e8a65231d99017001c56f0ba2c37749f4f1, delta 37c8044..65b51e8 and F1-F4 f6e6208..65b51e8): B1 is closed. No code path, docstring, help text, SKILL line or CHANGELOG line still says an old sentinel means "no other run" or "no re-collection"; another name's sentinel always gets the conditional. L2 is closed: A1 is killed by each TZ parameter on its own. L1 is only partly closed. Two wording fixes are recommended before merge, and neither blocks.
3. Head f74f4cb adds only those two wording fixes (L1', L3) and nits N1 to N3, text and comments, no behaviour change; the reviewer said they can follow without another code round. That commit was not re-reviewed.

Tests: tests/sudo_postinstall/test_secrets_collect.py, 55 passed on f74f4cb (baseline 44 on PR 5), through bg. Mutants were killed one at a time (11 in round 6, including a naive-local timezone parse killed at TZ=UTC and TZ=America/New_York, and the production lock path reading $TMPDIR). No real clipboard, vault or lock file was touched.

CHANGELOG: two lines added to the existing --yes entry under Unreleased.

🤖 Generated with Claude Code

---

Note

Medium Risk
Changes agent-facing clipboard collection semantics and exit codes; mishandling could leave secrets on the pasteboard or mis-route re-collection after parallel runs.

Overview
secrets-collect.py (--clipboard --yes) tightens agent-safe clipboard handling: failed pbcopy after a successful store (or dry-run) now exits 3 with an explicit “NOT cleared” line, while other failure modes keep 1/2/130.

Clipboard sentinels now stamp date, time, and zone (with backward-compatible parsing of the old HH:MM:SS form), and other-name errors add informational age (“N s/min ago”). Re-collection guidance for a different secret’s sentinel stays always conditional—age never implies the race is harmless; if the operator may have copied for this name before the stamp, both names must be re-collected.

SKILL.md and the GitHub recipe no longer tell agents to run printf 'cleared' | clip after every run; manual overwrite is only for exit 3, NOT cleared, no output, or giving up after a run that never read the clipboard.

Tests add coverage for dated sentinels, timezone-aware age math, exit 3 on clear failure, and that the default lock lives under ~/Library/Caches, not $TMPDIR.

^{Reviewed by Cursor Bugbot for commit 79c0939. Bugbot is set up for automated code reviews on this repo. Configure here.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-26T06:25:31.691416Z	f74f4cb	PR opened
🔒 Security Review	✅ Completed 2026-10-26T06:29:03.040222Z	f74f4cb	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: 13438b3a-97ab-4da5-babc-c42f1aefb646)

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: 9833b36a-7361-4d2a-ab4c-b7c784986595)

[linear-code]

VIB-119