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 Rohit-ATS/Airlock --skill access-reviewgit clone --depth 1 https://github.com/Rohit-ATS/AirlockWrote 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/rohit-ats/airlock/access-review)<a href="https://agentmods.dev/skills/rohit-ats/airlock/access-review"><img src="https://agentmods.dev/badge/skills/rohit-ats/airlock/access-review/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/rohit-ats/airlock/access-review"><img src="https://agentmods.dev/badge/skills/rohit-ats/airlock/access-review.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 | $0.00033 | $0.01849 |
| Opus 5 | $0.00016 | $0.00924 |
| Sonnet 5 | $0.00007 | $0.00370 |
| Haiku 4.5 | $0.00003 | $0.00185 |
Grade A, and why
access-review 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 9d 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 — 201 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Access review
A policy document and an effective permission are not the same thing, and the gap between them is where breaches live. This skill is about closing that gap before a human is asked to sign anything.
The rule that governs everything below: never report what a policy says. Report what a principal can do, computed by simulation against a shadow copy of the account. If you cannot simulate it, say you could not, and do not guess.
1. The three questions
For every principal in the change:
- What can they reach after this that they could not reach before? The delta, not the total. An approver cannot judge a 400-line policy document; they can judge four new verbs.
- What is the worst thing reachable through this grant? Not the intended use — the
maximum. Read access to
public.*includes the table nobody remembered was there. - When does it end? A grant with no expiry is refused by policy. There is no such thing as temporary access that nobody scheduled to remove.
2. Effective permissions, not declared ones
Four mechanisms make the declared policy an underestimate. Check each.
Transitive role membership
-- Postgres: everything prod_reader actually inherits, transitively
WITH RECURSIVE m AS (
SELECT oid, rolname FROM pg_roles WHERE rolname = 'prod_reader'
UNION
SELECT r.oid, r.rolname
FROM pg_auth_members am
JOIN pg_roles r ON r.oid = am.roleid
JOIN m ON m.oid = am.member
)
SELECT rolname FROM m;
A role granted a role granted a role is three documents away from the grant you are
reviewing. \du shows one level. Recurse.
PUBLIC grants
GRANT ... TO PUBLIC is invisible in a per-role audit and applies to the new principal
immediately.
SELECT table_schema, table_name, privilege_type
FROM information_schema.role_table_grants
WHERE grantee = 'PUBLIC';
Default privileges
ALTER DEFAULT PRIVILEGES means the grant covers tables that do not exist yet. This is
the one people miss: the review passes, and six weeks later a new table containing card
data is readable by a role nobody re-reviewed.
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.
- 9d ago First seen · 201 lines · 33 tokens per session scan A 1f3d298000a2
access-review is a skill published in the GitHub repository Rohit-ATS/Airlock (1 stars, last pushed yesterday), licensed MIT. It adds 33 tokens to every session and 1,849 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 2026-08-31.
Other skills, from other repositories
reweave-operator
Operate a Reweave self-healing web-data fleet — run pipelines, triage structural-drift incidents, review repair proposals, and manage the human approval workflow. Use whenever a monitored source is broken, a heal is pending, or the operator asks about pipeline health or impact.
A set of instructions for working with PDF files, which are documents designed to preserve their layout across devices.
pptx
A set of instructions for working with PowerPoint presentation files, including extracting content, editing slides, and creating decks.
oma-hwp
Convert HWP / HWPX / HWPML files to Markdown using kordoc. Extracts text, headings, tables, lists, images, footnotes, and hyperlinks. Use for Korean word processor files (Hangul), government documents, and AI-ready data preparation.
oma-pdf
Convert PDF files to Markdown using opendataloader-pdf. Extracts text, tables, headings, lists, and images with correct reading order. Use for PDF parsing, PDF to Markdown conversion, document extraction, and AI-ready data preparation.
gentle-ai-work-unit-commits
Plan commits as reviewable work units. Trigger: implementation, commit splitting, chained PRs, or keeping tests and docs with code.