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/AgentToolkit/altk-evolveWrote 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/commands/agenttoolkit/altk-evolve/adversarial-review)<a href="https://agentmods.dev/commands/agenttoolkit/altk-evolve/adversarial-review"><img src="https://agentmods.dev/badge/commands/agenttoolkit/altk-evolve/adversarial-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/commands/agenttoolkit/altk-evolve/adversarial-review"><img src="https://agentmods.dev/badge/commands/agenttoolkit/altk-evolve/adversarial-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 | $0.00026 | $0.01673 |
| Opus 5.5 | $0.00010 | $0.00669 |
| Sonnet 5.5 | $0.00005 | $0.00335 |
| Haiku 4.5 | $0.00003 | $0.00167 |
Grade A, and why
adversarial-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 29d 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 — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run an adversarial code review of pull request #$1, following docs/adversarial-review.md.
Method — evidence over opinion; every finding carries a reproduction or it doesn't ship. Do NOT report style nits.
Arguments
Parse $ARGUMENTS positionally and by flag, independently of each other:
$1— the PR number (required).- Flags — any argument starting with
--, in any position:--post(post without a second confirmation),--sandbox/--no-sandbox(force or skip container isolation; see step 3). - Repo override — the first non-flag argument after
$1, if there is one. Do not treat$2as the repo without checking it first:/adversarial-review 289 --posthas no repo override. Default<REPO>to this repo'supstreamremote, falling back toorigin.
Steps
-
Fetch & size the PR.
gh pr view $1 --repo <REPO> --json title,author,state,isCrossRepository,headRepositoryOwner,headRefName,headRefOid,baseRefName,baseRefOid,mergeable,mergeStateStatus,body,additions,deletions,changedFiles,commitsgh api repos/<REPO>/pulls/$1 --jq .author_association— not agh pr viewfield; asking for it there fails the whole call.gh pr view $1 --repo <REPO> --json files --jq '.files[]|"\(.additions)+ \(.deletions)- \(.path)"' | sort -rnto see the shape.- Record
baseRefOidandheadRefOid— every diff and baseline below uses those two SHAs, never a hard-codedmain. A PR may target a release or feature branch; diffing against the wrong base misattributes pre-existing defects to the PR, or hides its changes entirely. - If the PR body claims specific bugs/fixes, note them — they become verification targets.
-
If this is a re-review, load every prior finding. The status table in step 7 is per finding, so the summary bodies alone are not enough:
gh api --paginate repos/<REPO>/pulls/$1/reviews— review bodies and verdicts.gh api --paginate repos/<REPO>/pulls/$1/comments— the individual inline findings.- Correlate each inline comment to its review via
pull_request_review_id, and keeppath+line/original_line. Every prior finding must come back with a fixed / still-open / regressed verdict.
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.
- 29d ago First seen · 58 lines · 26 tokens per session scan A 7b109e6933b7
adversarial-review is a command published in the GitHub repository AgentToolkit/altk-evolve (124 stars, last pushed today), licensed Apache-2.0. It adds 26 tokens to every session and 1,673 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-09-17.
Other commands, from other repositories
init
Install the formatters this repository needs, with every command visible before it runs.
git
The pre-finish status: branch, hygiene findings, message checks, workflow lint, template state.
extend
Capture a mid-PR sub-idea and implement it onto the current open PR's branch — no new branch, no new PR. Preserves Verify → Review → Deliver.
esp-harden
Harden and inspect ESP32 firmware for field failures, crashes, memory, and security.
security-review
Depth-first security audit of the current git diff — OWASP Top 10, secret leakage, insecure-by-default APIs, and language-specific foot-guns. Pass a PR number to audit a specific PR's diff.
harness-onboarding
Generate a human-readable onboarding document from HARNESS.md, AGENTS.md, and REFLECTIONLOG.md — a friendly guide for new team members.