When a PR modifies the admin settings surface, file/class structure, or data flow, the QA reviewer now flags the PR as a MAJOR concern unless the PR description references an accompanying update to the CTO app-map at `Peptide Repo CTO/app-maps/<repo-slug>.md`.
The CTO workspace is a separate repo, so CI cannot verify the doc update mechanically — the reviewer inspects the PR description for an explicit mention.
Why: the CTO-level app-maps are the single-pane-of-glass reference that lets the CTO answer questions like 'which settings does this plugin expose?' without grepping the repo each time. Without a merge-gate prompt, these maps drift out of sync with the code.
Co-authored-by: peptiderepo <peptiderepo@users.noreply.github.com>
Neurometric markets a task-specialized PR-Summarizer SLM (Qwen3-4B +
specialist system prompt) on a free tier. Adding it as a 4th voice in
the existing A/B reviewer panel to see whether a specialized small
model catches things the 3 general-purpose voices miss.
The SLM does not honor OpenAI tool-calling, so it cannot return our
structured submit_review schema. Wired as a "prose voice":
- MODELS tuple gains a 6th provider_type field
("openrouter_tools" | "neurometric_prose")
- call_model() dispatches; new call_neurometric_prose() path hits
api.neurometric.ai/v1 without tools, embeds raw markdown output,
and infers verdict from keyword heuristic
- Render branch in PR comment marks the verdict as heuristic and
notes tokens/cost are not tracked (free tier)
- Never the primary voice; heading still driven by Haiku
Also hardens the shared render path against models that omit the
optional 'summary' field (rev.get('summary') fallback).
Requires new repo secret NEUROMETRIC_API_KEY (already set on all four
peptiderepo repos). Voice self-skips with inline error if missing.
Agent-Session: cowork-neurometric-qa-voice-20260415
Co-authored-by: peptiderepo <peptiderepo@users.noreply.github.com>
* ci: cost cap + 3-model A/B harness on CTO reviewer
Two enhancements to the PR reviewer:
1. Monthly spend cap: tracked in Repo Variables (CTO_REVIEW_MONTH,
CTO_REVIEW_SPEND_CENTS, CTO_REVIEW_CAP_CENTS). Default cap $10/month.
Auto-resets on month rollover. Posts a skip-comment on PRs when cap hit.
2. A/B harness: reviews every PR with 3 models in parallel
(Claude Haiku 4.5, DeepSeek V3.1, Kimi K2 Thinking). Comment shows
all 3 verdicts side-by-side so we can compare quality/cost on real
PRs before picking a permanent default. Drop to 1 model by editing
the MODELS list after A/B window closes.
Agent-Session: cowork-reviewer-ab-20260414
* ci: use CTO_ADMIN_TOKEN for repo variables + restore formal APPROVE/REQUEST_CHANGES
Default GITHUB_TOKEN cannot mutate Repository Variables (always 403).
Use a PAT-backed secret CTO_ADMIN_TOKEN for the budget counter steps.
Also restores formal APPROVE/REQUEST_CHANGES verdicts now that repo
setting can_approve_pull_request_reviews is enabled.
Agent-Session: cowork-reviewer-ab-20260414
* ci: address review feedback — per-model timeout + simpler cost math
- Per-model timeout of 60s so one hung model does not block siblings
- Simplify cost_cents math (cost_usd * 100 directly, no double-rounding)
Defer: fine-grained PAT for CTO_ADMIN_TOKEN (requires manual OAuth device flow).
Agent-Session: cowork-reviewer-ab-20260414
* ci: stop exposing CTO_ADMIN_TOKEN as env var (security cleanup)
Pass the PAT only via github-token param. continue-on-error on the
budget step so a missing/expired secret degrades gracefully instead
of failing the whole workflow.
Agent-Session: cowork-reviewer-ab-20260414
* ci: post CTO review as COMMENT instead of APPROVE
GITHUB_TOKEN cannot formally approve PRs without a repo setting toggle.
Verdict is still prominent in the first line of the review body, and
the human merges based on reading it — no behavior loss.
Agent-Session: cowork-reviewer-fix-20260414
* fix(ci): ssh-keyscan failure no longer fails the deploy
ssh-keyscan was being called with set -e in effect and stderr suppressed,
so a single keyscan failure killed the entire deploy with no diagnostic
output. Since rsync/ssh use StrictHostKeyChecking=no, the known_hosts
entry is not load-bearing — making the keyscan best-effort is safe.
stderr is now captured to /tmp/keyscan.err and emitted as a warning so
future failures are debuggable.
Agent-Session: cowork-deploy-fix-20260414
* ci: add main-push audit tripwire
Flags direct pushes to main that did not come from a merged PR.
Free-plan workaround since branch protection requires GitHub Pro.
* docs: add Git Workflow section to CONVENTIONS.md
Documents the PR-gated review process enforced at the agent layer.
Free-plan workaround since branch protection requires GitHub Pro.
* ci: add Claude-powered CTO PR review workflow
Auto-reviews every PR with Claude Haiku 4.5 and posts a structured
review (APPROVE / REQUEST_CHANGES / COMMENT). Does NOT auto-merge — a
human still clicks Merge. Requires ANTHROPIC_API_KEY secret on the repo.
* ci: switch CTO reviewer from Anthropic direct to OpenRouter
Reuses the OPENROUTER_API_KEY secret that already backs the other AI
plugins. Consolidates AI spend on one bill. Model: anthropic/claude-haiku-4.5.
