Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add bop-clocktower/canary --skill canary-pr-guardiangit clone --depth 1 https://github.com/bop-clocktower/canaryWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/bop-clocktower/canary/canary-pr-guardian)<a href="https://agentmods.dev/skills/bop-clocktower/canary/canary-pr-guardian"><img src="https://agentmods.dev/badge/skills/bop-clocktower/canary/canary-pr-guardian/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/bop-clocktower/canary/canary-pr-guardian"><img src="https://agentmods.dev/badge/skills/bop-clocktower/canary/canary-pr-guardian.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00088 | $0.01419 |
| Opus 5 | $0.00044 | $0.00709 |
| Sonnet 5 | $0.00018 | $0.00284 |
| Haiku 4.5 | $0.00009 | $0.00142 |
Grade A, and why
canary-pr-guardian scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 130 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Canary: PR Guardian
Guards a change's test quality before it lands. Composes the deterministic
Tier-0 diff-coverage engine with two native agents — canary-test-reviewer
(read-only audit) and canary-test-author (authoring) — under a strict
write-safety model. This is the Option A driver: the engine never calls an
LLM; this skill invokes the agents in-session and enforces
stage-and-block-once.
On the tier numbers. Here they are the values of
canary guardian pr-check --tier 0|1|2, not a repo-wide capability scale — Tier
1 means "the --tier 1 pass," which is the agent audit. Tier-0 is the one
number with a repo-wide meaning (deterministic, no network, no agent), and
Tier-1/Tier-2 are guardian-local by
ADR 0015.
Do not carry them into other skills.
When to Use
- Before opening or updating a PR, to check that new/changed code is tested.
- As a pre-commit companion when
preCommit.authorTests: trueis set and you want the guardian to author the missing tests for you (at the desk only). - NOT in CI for Tier-2 write-back — that is a NON-GOAL. CI runs Tier-0 only (the
CANARY_GUARDIAN_AGENTenv is unset there).
Safety model (non-negotiable)
- NEVER commit or push. This skill only authors and
git adds. The human reviews the staged tests and re-commits. - Honor every
skippedreason fromauthor-planverbatim (opt-in-off / tier / fork / collision / loop-guard). Never override a skip. - Block once. When
block.block == true, print the block message and stop — leave the staged tests for the human. The loop-guard sentinel you write in Phase 3 is what stops the guardian re-authoring over its own output on the next run. It is stamped with the currentHEADand expires by itself once the human's review commit movesHEAD— never delete it yourself. - Authoring is opt-in. No
preCommit.authorTests: true⇒ no writes, ever.
Phases
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago Changed · +8 lines 9d66e5db147e
- 11d ago First seen · 122 lines · 88 tokens per session scan A c52fa351286c
canary-pr-guardian is a skill published in the GitHub repository bop-clocktower/canary (4 stars, last pushed today), licensed MIT. It adds 88 tokens to every session and 1,419 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
adk-go-self-review
Review an ADK Go change the way a maintainer will — a fresh-context pass over the whole diff, five lenses (correctness and tests, scope, simplicity, style, adk-python parity), and the mutation check that proves your tests pin the change. Use before opening a PR, before any later push that changes code, and when asked…
gentle-ai
Use Gentle AI harness discipline for Pi work: clarify first, preserve OpenSpec artifacts, use strict TDD where available, delegate through subagents when useful, and protect review workload.
software-test-review
Evaluate the quality of TDD tests against slice acceptance criteria, codebase conventions, and Red-phase execution results, producing a structured review with Accept or Revise recommendations. Use when tests written during the Red phase of red-green-refactor need quality review — checking coverage of acceptance…
orchestrated-execution
Execute work units through the rigorous 4-phase Metaswarm cycle (Implement -> Validate -> Adversarial Review -> Commit) with independent quality gate enforcement.
software-code-refactoring
Improve production code quality while preserving all existing test behavior. Commonly used for the Refactor phase of TDD red-green-refactor, but applicable to any codebase with tests. Use when production code works but needs cleanup — reducing duplication, improving naming, simplifying complexity, aligning with…
quality-playbook
Run a complete quality engineering audit on any codebase. Derives behavioral requirements from the code, generates spec-traced functional tests, runs a three-pass code review with regression tests, executes a multi-model spec audit (Council of Three), and produces a consolidated bug report with TDD-verified patches.…