Borrowing it
Nothing to install: this file belongs to wbrown/janus-datalog. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/wbrown/janus-datalog/main/.claude/skills/comment-sweep/SKILL.mdgit clone --depth 1 https://github.com/wbrown/janus-datalogWrote 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/wbrown/janus-datalog/comment-sweep)<a href="https://agentmods.dev/skills/wbrown/janus-datalog/comment-sweep"><img src="https://agentmods.dev/badge/skills/wbrown/janus-datalog/comment-sweep.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00154 | $0.02235 |
| Opus 5 | $0.00077 | $0.01118 |
| Sonnet 5 | $0.00031 | $0.00447 |
| Haiku 4.5 | $0.00015 | $0.00224 |
Grade A, and why
comment-sweep 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 8d 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 — 176 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Comment Sweep
A comment earns its place by telling a reader of this file, as it is now something the code does not already say. Two defects break that, and they need hunting together because the first decays into the second.
The three targets, co-equal
1. Written for a diff reader. The audience is someone comparing this file against its previous version, or against an alternative the author considered and rejected. Neither is in the file.
2. False about the present. The comment states something about current behaviour that is not true.
3. True, present-tense, and four times too long. Every sentence carries something; together they argue where one would inform.
The first two are not independent. A comment about the present gets re-read every time someone edits the code beneath it, so it gets corrected. A comment about the past never does — nothing rechecks it, and it drifts into falsehood unopposed. Target 1 is therefore the prophylactic for target 2, and target 2 is the more damaging of the pair: diff narration is noise a reader skips, a false comment is followed. Target 3 is where the other two hide: a rejected alternative is easiest to smuggle in as the fourth sentence of something that opened with a real invariant.
A sweep that hunts only target 1 will find the others and walk past them. That has happened; brief for all three explicitly.
Target 1: forms to remove
- Prior state of the code, the test, or a previous comment: "no longer", "used to be", "previously", "this read X", "was renamed", "(original implementation)", "a wrapper existed here and was removed", "until ".
- Counterfactual justification of a past change: "would break", "would have", "could not tell X from Y", "still passes", "which is how X survived".
- Session or review evidence as motivation: commit SHAs, before/after measurements ("from 217 allocations to 240"), "observed and corrected ".
- Narration of the work rather than the code: "Caught by TestX", "found while implementing", "the earlier version conflated", "added in response to review".
- Self-correcting scratch work left in place: "Actually, looking at it again...", "Wait, let me re-read the semantics...".
- A count of things in OTHER files offered as justification: "all four production sites", "the eleven producers". A count about this file's own data is fine.
- On an assertion: why THIS ASSERTION FORM was chosen, rather than what the expected value means. But see the carve-out below — this one has a sharp edge.
- A downstream application or workload name. Session-private, and this repo is public. The bug it names is real provenance; state the defect and its symptom, drop the name.
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.
- 8d ago First seen · 176 lines · 154 tokens per session scan A c87f8f0fdcdb
comment-sweep is a skill published in the GitHub repository wbrown/janus-datalog (51 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 154 tokens to every session and 2,235 once invoked, about $0.0008 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
autoreview
Pre-commit/ship code review: Codex default; optional Claude or Pi.
rework-rate
Measure and interpret PR rework rate — the emerging 5th DORA metric.
omh-code-review
This is a Hermes-native code-review workflow skill.
revdiff-plan
Review the last Codex assistant message (plan, analysis, or proposal) with inline annotations in a TUI overlay. Extracts the most recent response from Codex rollout files and opens it in revdiff for review and annotation. Activates on "revdiff-plan", "review plan with revdiff", "annotate plan", "review last response"…
code-reviewer
Code review specialist focused on patterns, bugs, security, and performance.
agent-teams-simplify-and-harden
Implementation + audit loop using parallel agent teams with structured simplify, harden, and document passes. Spawns implementation agents to do the work, then audit agents to find complexity, security gaps, and spec deviations, then loops until code compiles cleanly, all tests pass, and auditors find zero issues or…