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 JKHeadley/instar --skill iterative-converging-auditgit clone --depth 1 https://github.com/JKHeadley/instarWrote 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/jkheadley/instar/iterative-converging-audit)<a href="https://agentmods.dev/skills/jkheadley/instar/iterative-converging-audit"><img src="https://agentmods.dev/badge/skills/jkheadley/instar/iterative-converging-audit/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/jkheadley/instar/iterative-converging-audit"><img src="https://agentmods.dev/badge/skills/jkheadley/instar/iterative-converging-audit.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00000 | $0.02270 |
| Opus 5 | $0.00000 | $0.01135 |
| Sonnet 5 | $0.00000 | $0.00454 |
| Haiku 4.5 | $0.00000 | $0.00227 |
Grade A, and why
iterative-converging-audit 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 10d 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 — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.
iterative-converging-audit — Audit to Convergence, Not to Exhaustion-of-Patience
A single audit pass is never thorough. The first sweep has blind spots; the fixes themselves reveal new instances or introduce new ones; and "I looked once and stopped finding things" usually means "I got tired," not "there is nothing left." The only honest definition of a complete audit is a converged one: a re-run that finds zero new discoveries.
This skill turns that into a structurally enforced loop. It applies to ANY "find all instances of X" task — security audits, safety audits, code reviews, research sweeps, compliance checks, dependency audits, dead-code hunts, "are there other places we do this wrong" investigations.
This skill enforces the Iterative Audit to Convergence constitution standard (docs/STANDARDS-REGISTRY.md). You do not declare an audit complete on round 1. You track rounds. You re-audit after every fix batch.
When to Activate This Skill
- Any request to "find all", "audit", "sweep", "review everything", "make sure we got everything".
- After fixing a bug, when the same class of bug likely exists elsewhere ("where else do we do this?").
- A security or safety audit where a missed instance is dangerous.
- A research/review question where one source or one angle is not enough.
- Whenever you catch yourself about to say "I checked, looks clean" after a SINGLE pass.
The Loop (do not skip steps)
Step 0 — Frame the audit (write this down first)
Before sweeping, define — explicitly, in the audit ledger:
- Target pattern — what exactly are you finding? Be precise. "Silent LLM fallbacks in gating paths", not "bad error handling".
- Search surface — where could instances live? List the angles/locations. The first list is always incomplete; you will add to it.
- Classification rule — for each finding, what are the buckets? (e.g. DANGEROUS / MITIGATED / ADVISORY-OK). A finding is not "done" until it is classified.
- Fix policy — what does "fixed" mean per bucket? (remediate / annotate-as-accepted-with-reason / escalate).
- Convergence criterion — usually: a full re-sweep surfaces zero findings not already in the ledger. State it concretely.
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.
- 10d ago First seen · 102 lines · 0 tokens per session scan A 43de6e834a3b
iterative-converging-audit is a skill published in the GitHub repository JKHeadley/instar (79 stars, last pushed yesterday), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,270 tokens. 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
code-review
Use when reviewing a code change or diff for correctness, security, missing tests, and convention violations before opening or approving a PR. Review independently and adversarially, then fix high-confidence issues.
simplify
Use after writing or modifying code to improve its clarity, reuse, simplicity, and efficiency without changing behavior. Not for finding correctness bugs — that is code review.
review-implement-phase
Implements triaged review actions, commits focused fixes, and posts Done plus resolves threads. Use when the user wants only the implementation phase of the review-framework workflow.
no-bare-casts
Writing as in TypeScript or TSX production code, modifying a file that contains a bare as cast, silencing a type error with a cast, encountering as unknown as, or reviewing a cast site.
agent-code-review-swarm
Agent skill for code-review-swarm - invoke with $agent-code-review-swarm.
update-pr
Update the pull request for the current session. Use when the user wants to push new changes to an existing PR.