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 bostonaholic/team --skill pr-watch-as-reviewergit clone --depth 1 https://github.com/bostonaholic/teamWrote 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/bostonaholic/team/pr-watch-as-reviewer)<a href="https://agentmods.dev/skills/bostonaholic/team/pr-watch-as-reviewer"><img src="https://agentmods.dev/badge/skills/bostonaholic/team/pr-watch-as-reviewer/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/bostonaholic/team/pr-watch-as-reviewer"><img src="https://agentmods.dev/badge/skills/bostonaholic/team/pr-watch-as-reviewer.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 | $0.00037 | $0.00907 |
| Opus 5 | $0.00018 | $0.00453 |
| Sonnet 5 | $0.00007 | $0.00181 |
| Haiku 4.5 | $0.00004 | $0.00091 |
Grade A, and why
pr-watch-as-reviewer 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 yesterday.
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 — 81 lines — stays where its author put it; the contents beside it link to each section on GitHub.
pr-watch-as-reviewer — reviewer-side watch-and-approve loop
pr-watch-as-reviewer is the reviewer-side mirror of
pr-watch-as-author. You post
review comments on a PR you are reviewing, then arm the skill. It polls
until every piece of feedback you left is settled, re-reviews each
settlement on substance as it lands, and only when every settlement
passes casts gh pr review --approve on your behalf and stops. Model
invocation is disabled (disable-model-invocation: true): on a PR with
auto-merge enabled, an approval can transitively trigger an irreversible
merge, so only a deliberate human invocation arms the watch.
agents/openai.yaml restates the same guard for Codex as
policy.allow_implicit_invocation: false.
Feedback comes in two shapes, and the watch tracks both:
- a review thread — an inline comment anchored to a diff line, which GitHub gives a resolved/unresolved bit.
- a plain PR comment — a top-level issue comment on the conversation tab, which GitHub gives no resolution bit at all. A whole-PR review posted as one comment body (the common shape for an automated or summary review) lands here.
That asymmetry drives the whole design below. A thread has an explicit author action — resolving it — that says "I am done with this". A plain comment has no such affordance: there is nothing for the author to click.
Neither is trusted on its own. The only thing that settles either is the state of the branch, read as it now stands. A resolve is a claim by the person whose code you are approving; it can be clicked over a concern that was never addressed. So every item is verified against the current code, always. The two shapes differ only in which way an unclear read falls:
- a plain comment requires that the head advanced after it — no push since the comment means nothing could have addressed it — and an unclear read leaves it unsettled.
- a resolved thread is verified too, but the author's explicit assertion earns deference: overturning it takes very high confidence and strong disagreement, not a quibble.
The approval body discloses how many approved items were of each shape, so a reader can see which evidence the approval rested on.
Every verdict is published where the author will see it. A reply that meets the concern resolves the thread. A reply that does not draws a rebuttal naming the specific gap. A reply that is read, judged, and then left sitting is the failure mode this skill exists to avoid: the author cannot tell a considered acceptance from an unread one, and a thread that stays open with no answer reads as a reviewer who disappeared. Silence is not an answer.
What ships with it
11 files 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.
- agents/openai.yaml 249 B
- references/01-hard-rules.md 5.0 KB
- references/02-input.md 2.4 KB
- references/03-execution.md 13 B
- references/04-1-arm.md 10 KB
- references/05-2-tracked-set-and-gate.md 4.1 KB
- references/06-3-bounded-cycle-mechanics.md 600 B
- references/07-4-poll.md 17 KB
- references/08-5-stop-conditions.md 2.4 KB
- references/09-6-approve.md 8.3 KB
- references/10-compaction-defense.md 4.1 KB
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.
- yesterday Changed · -10 tokens per session 6ccfbffd1758
- 7d ago Changed · -1,027 lines · -186 tokens per session fcff10e860ff
- 9d ago Changed · +181 lines · +29 tokens per session 75e60e8a55eb
- 12d ago First seen · 927 lines · 204 tokens per session scan A 5eb0c306014d
pr-watch-as-reviewer is a skill published in the GitHub repository bostonaholic/team (11 stars, last pushed today), licensed MIT. It adds 37 tokens to every session and 907 once invoked, about $0.0002 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 2026-08-30.
Other skills, from other repositories
codebase-sync
Convention discovery and rule generation from codebase analysis. Scans project structure, builds search indexes, identifies patterns, and generates enforceable rules.
code-review-patterns
Multi-dimensional code assessment across security, quality, performance, and maintainability with confidence-gated reporting (>=80%) and Router Contract generation.
code-review-orchestration
6-agent parallel code review orchestration covering architecture, security, performance, testing, quality, and documentation dimensions with weighted scoring.
code-review-pipeline
Multi-dimensional code review across correctness, security, performance, and maintainability with confidence-gated reporting and remediation loops.
orchestrated-execution
Execute work units through the rigorous 4-phase Metaswarm cycle (Implement -> Validate -> Adversarial Review -> Commit) with independent quality gate enforcement.
plan-implementation
Disciplined execution of approved plans with step-by-step verification, phase checkpoints, failure investigation, and mandatory code/security reviews.