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 mishahanin/heading-os --skill scrutinizegit clone --depth 1 https://github.com/mishahanin/heading-osWrote 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/mishahanin/heading-os/scrutinize)<a href="https://agentmods.dev/skills/mishahanin/heading-os/scrutinize"><img src="https://agentmods.dev/badge/skills/mishahanin/heading-os/scrutinize/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/mishahanin/heading-os/scrutinize"><img src="https://agentmods.dev/badge/skills/mishahanin/heading-os/scrutinize.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 | $0.00186 | $0.04433 |
| Opus 5 | $0.00093 | $0.02217 |
| Sonnet 5 | $0.00037 | $0.00887 |
| Haiku 4.5 | $0.00019 | $0.00443 |
Grade A, and why
scrutinize 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 — 271 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Scrutinize
Manually-invoked quality gate. It runs a maximum-effort VIIA pass (Validate - Identify - Improve - Adjust) over a target. An adversarial layer then filters the false positives. Each finding carries a concrete fix and a confidence score, and the gate blocks progress until the user approves the batch.
Principal-engineer posture: find what is wrong, not what works. Every finding needs evidence. Every BLOCKER and HIGH survives a debate first. No shortcut exits. Version history: references/version-history.md.
Variables
One optional target plus the flags in references/flags.md, which Phase 0
step 1 already loads. That catalog is the source of truth. The one-line copies
in argument-hint, x-heading-capability.how and x-heading-routing.label
carry the same flags. tests/test_three_flag_lists_that_described_one_skill.py
fails when a copy drifts from the catalog, or names a flag no code reads.
When to Engage
Manual invocation only.
Use it before approving a high-stakes plan, or after /implement to audit the changes against that plan. Use it on a file or a dir when something feels off, and periodically on the workspace to catch drift.
Do NOT use it for artifact grading (/evaluate), draft fact-checking (/validate), decision reasoning (/deep-think), or content-quality review of a post or an email (/evaluate).
Phase 0 - Context Loading
-
Load every reference file under
.claude/skills/scrutinize/references/. The directive is the directory, not a list: each file states what consumes it. -
Load applicable rules:
.claude/rules/{development-standards,hidden-chars,security,classification,voice}.md. -
Resolve target per
references/target-detection.md. Parse an explicit argument from the invocation when one is present. Otherwise apply the priority order plan > execution (git) > menu. Print a confirmation line for priorities 2 to 4 and wait for the user. Use the new target when they redirect you.
What ships with it
25 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.
- evals/benchmark.json 4.7 KB
- evals/cases/case-1-viia-structure.json 406 B
- evals/cases/case-2-severity-tags.json 367 B
- evals/cases/case-3-batched-approval.json 474 B
- evals/cases/case-4-disprove-step.json 695 B
- evals/cases/case-5-confidence-score.json 592 B
- evals/cases/case-6-flag-as-fp.json 841 B
- evals/cases/case-7-trajectory-target.json 738 B
- evals/cases/case-8-deterministic-audit.json 606 B
- references/approval-block.md 3.0 KB
- references/bias-mitigation.md 12 KB
- references/code-review-voice.md 4.5 KB
- references/eval-case-template.md 17 KB
- references/flags.md 1.9 KB
- references/observability.md 6.7 KB
- references/refutation-protocol.md 14 KB
- references/relentless-adaptive.md 6.3 KB
- references/report-format.md 3.9 KB
- references/role-lenses.md 7.2 KB
- references/severity-grid.md 7.0 KB
- references/target-detection.md 5.6 KB
- references/trajectory-evaluation.md 10 KB
- references/version-history.md 1.8 KB
- references/viia-framework.md 11 KB
- references/workspace-areas.md 6.3 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.
- 8d ago First seen · 271 lines · 186 tokens per session scan A 8810d174a189
scrutinize is a skill published in the GitHub repository mishahanin/heading-os (11 stars, last pushed 3d ago), licensed Apache-2.0. It adds 186 tokens to every session and 4,433 once invoked, about $0.0009 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 2026-09-03.
Other skills, from other repositories
review-loop
Run the adversarial verification loop — implement, then hand the change to a fresh checker that did not write it, fix what it finds, and re-dispatch until APPROVE. Use before claiming any behavioural change is done, and on requests like "review loop", "adversarial review", "independent review", "get this verified"…
ai-slop-cleaner
Post-implementation cleanup that removes AI-generated bloat while preserving functionality. Runs pass-by-pass with test verification after each pass. Activate after kraken/spark complete a feature, or when a codebase needs hygiene work.
dx-audit
Audits libraries, CLIs, and SDKs using 38 rules for public contracts, package exports, piped output, errors, and configuration. Use when asked to "audit my CLI", "review my SDK", "make this agent-friendly", or diagnose package type resolution. For agentic product trust use ax-audit; for docs use docs-writing.
anti-stub-scan
Scan a diff for stub patterns and TODO markers.
bug-hunt
Proactive bug-hunting workflow. Assesses codebase risk through complexity, coverage, and structural analysis, then spawns focused investigators that write reproducing tests to validate suspected bugs. Thoroughness over speed. Advisory only — produces findings and proposes tickets; does not implement fixes.
gs-review
Pre-landing PR review - gstack staff-engineer code review army. CARL TRIGGERS: review the code, audit this branch, check this PR, find bugs, code review, review my changes. SOURCE: garrytan/gstack/review, integrated as gs-review on 2026-05-29.