Re-opened after a phantom-completion audit. A verify-first judge previously closed the original issue as ALREADY_IMPLEMENTED with zero code change; an independent adversarial re-audit against current main confirmed the work is still missing.
Audit finding (files: patrol.py)
patrol.py: Independently confirmed phantom completion. Issue #119 requests a patrol that detects/proposes Codex (GPT) model updates. The cited patrol.py:389 is inside call_ai(), which builds the Anthropic/GLM /v1/messages request payload (json.dumps({"model", "max_tokens", "messages"})) — entirely unrelated to Codex model freshness. grep -i "codex|gpt" over patrol.py returns zero hits. The registered patrols are patrol_code_review, patrol_research, patrol_iteration (run_patrol loop ~line 2158) plus patrol_org_coherence and patrol_cross_repo_deps (main); none check Codex fallback models. patrol_research (lines 574-651) only inspects npm/bun outdated deps and Dependabot security alerts — no model-version freshness check. The Codex fallback models are still hardcoded in worker.py:25-26 (FALLBACK_CODEX_NORMAL_MODEL="gpt-5.5", FALLBACK_CODEX_DOCS_MODEL="gpt-5.4-mini"; also used in claude_ops.py:123,125) with no update-detection patrol anywhere. The feature does not exist; the close as ALREADY_IMPLEMENTED with an irrelevant citation and zero code change is a phantom completion.
Note
Some original citations were fabricated (line numbers that do not exist in the file) — that is itself evidence the close was not based on real code. Please verify against current main.
Filed by cross-repo phantom-completion audit (from claude-bot-go operator). Reported as an issue for the maintainers; not auto-implemented. Checked for duplicates before filing.
Re-opened after a phantom-completion audit. A verify-first judge previously closed the original issue as ALREADY_IMPLEMENTED with zero code change; an independent adversarial re-audit against current
mainconfirmed the work is still missing.Audit finding (files: patrol.py)
patrol.py: Independently confirmed phantom completion. Issue #119 requests a patrol that detects/proposes Codex (GPT) model updates. The cited patrol.py:389 is inside call_ai(), which builds the Anthropic/GLM /v1/messages request payload (json.dumps({"model", "max_tokens", "messages"})) — entirely unrelated to Codex model freshness. grep -i "codex|gpt" over patrol.py returns zero hits. The registered patrols are patrol_code_review, patrol_research, patrol_iteration (run_patrol loop ~line 2158) plus patrol_org_coherence and patrol_cross_repo_deps (main); none check Codex fallback models. patrol_research (lines 574-651) only inspects npm/bun outdated deps and Dependabot security alerts — no model-version freshness check. The Codex fallback models are still hardcoded in worker.py:25-26 (FALLBACK_CODEX_NORMAL_MODEL="gpt-5.5", FALLBACK_CODEX_DOCS_MODEL="gpt-5.4-mini"; also used in claude_ops.py:123,125) with no update-detection patrol anywhere. The feature does not exist; the close as ALREADY_IMPLEMENTED with an irrelevant citation and zero code change is a phantom completion.
Note
Some original citations were fabricated (line numbers that do not exist in the file) — that is itself evidence the close was not based on real code. Please verify against current
main.Filed by cross-repo phantom-completion audit (from claude-bot-go operator). Reported as an issue for the maintainers; not auto-implemented. Checked for duplicates before filing.