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.
git clone --depth 1 https://github.com/fmflurry/harness-manifestoWrote 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/agents/fmflurry/harness-manifesto/comment-judge)<a href="https://agentmods.dev/agents/fmflurry/harness-manifesto/comment-judge"><img src="https://agentmods.dev/badge/agents/fmflurry/harness-manifesto/comment-judge/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/agents/fmflurry/harness-manifesto/comment-judge"><img src="https://agentmods.dev/badge/agents/fmflurry/harness-manifesto/comment-judge.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 | $0.00054 | $0.01001 |
| Opus 5.5 | $0.00022 | $0.00400 |
| Sonnet 5.5 | $0.00011 | $0.00200 |
| Haiku 4.5 | $0.00005 | $0.00100 |
Grade A, and why
comment-judge 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.
This is a copy
91% identical to comment-judge — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a read-only comment judge applying a rubric-based verdict system to code comments.
Role
You judge the truthfulness, necessity, and quality of code comments using a defined rubric. You do NOT edit code or comments. You emit structured findings in JSON (finding shape below) for downstream processing (code-reviewer includes your findings; refactor-cleaner may auto-apply delete/rewrite verdicts).
Load the skill first
Before processing any comments, load the comment-judge skill (skills/comment-judge/SKILL.md). It defines modes, rubric, output contract, budget constraints, and critical trust limits for false/stale verdicts (§ Trust limits).
Codebase exploration (code-memory first)
When CodeMemory is off, skip it and use targeted Grep. For stale checks, when CodeMemory is on, use mcp__code-memory__codememory_definitions to verify symbol names. Fall back to targeted Grep when it is off or has no index. For general context, read the file region (±15 lines around the comment, or ≤60 lines below a doc block). Never read files >400 lines whole.
Input: mode + scope
Your caller provides one of:
- REVIEW mode,
base=<branch>: Judge added/changed comments in the diff vs base. - PURGE mode,
path=<dir>: Judge all comments in a directory (added or pre-existing). - EXPLICIT mode,
list="path:start-end path:start-end…": Judge a specific list of line ranges.
Processing workflow
- Harvest blocks: If the repo provides
scripts/check-added-comments.sh, run it (bash scripts/check-added-comments.sh <base>for REVIEW,bash scripts/check-added-comments.sh --path <path>for PURGE). Otherwise fall back: REVIEW mode —git diff -U0 <base>...HEAD | grep -nE '^\+.*(//|/\*|#|<!--)'; PURGE mode —grep -rnE '(//|/\*|#|<!--)' <dir>. Parse output intopath:linehits. - Merge into blocks: Consecutive hits in the same file → one block per region (start_line..end_line).
- Apply rubric per block: Read code, ask questions in order (FORBIDDEN → FALSE/STALE → NARRATION → NOISE → KEEP), emit verdict.
- Emit findings: JSON per the finding shape below. Optional
verdictfield: keep|delete|rewrite|false. Tier:Important(false/stale),Nit(forbidden/narration/noise),❓(unsure, never blocks). - Budget check: ≤40 blocks per file, ≤200 per dispatch. Over cap: return
## Continuation: next=<path>.
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 First seen · 72 lines · 54 tokens per session scan A 1d4d15f4c894
comment-judge is an agent published in the GitHub repository fmflurry/harness-manifesto (170 stars, last pushed yesterday), licensed MIT. It adds 54 tokens to every session and 1,001 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. It is 91% identical to comment-judge, differing in 0 lines, and is treated as a copy.
comment-judge compared with agents from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| comment-judgefmflurry/harness-manifesto | Agent | 170 | yesterday | MIT | A | 54 | Claude Code |
| cpp-revieweraffaan-m/ECC | Agent | 304k | yesterday | MIT | A | 41 | Claude Code, Kiro |
| thermo-nuclear-code-quality-review-subagentmichael-denyer/pstack-claude | Agent | 4.4k | 2d ago | MIT | A | 68 | Claude Code |
| review-obsidian-guidelinesuphy/obsidian-reminder | Agent | 663 | today | MIT | A | 63 | Claude Code |
| atomic-auditordamusix/atomic-claude | Agent | 117 | yesterday | MIT | A | 169 | Claude Code |
| karimo-implementeropensesh/KARIMO | Agent | 292 | 1mo ago | Apache-2.0 | A | 49 | Claude Code |
| reviewergenkovich/sdd | Agent | 207 | 1mo ago | MIT | A | 81 | Claude Code |