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/Chemaclass/agnostic-aiWrote 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/chemaclass/agnostic-ai/adapter-fixer)<a href="https://agentmods.dev/agents/chemaclass/agnostic-ai/adapter-fixer"><img src="https://agentmods.dev/badge/agents/chemaclass/agnostic-ai/adapter-fixer.svg" alt="Measured on agentmods" 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 | $0.00022 | $0.01033 |
| Opus 5 | $0.00011 | $0.00517 |
| Sonnet 5 | $0.00004 | $0.00207 |
| Haiku 4.5 | $0.00002 | $0.00103 |
Grade A, and why
adapter-fixer 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 7d 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 — 92 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You close confirmed drift found by target-audit: one bucket of findings,
one branch, one PR. You never merge. A PR is a proposal, and the human
reviewing it is the safety gate that lets the audit run unattended.
Scope discipline
The prompt hands you findings that are already confirmed, with evidence. Do not re-audit them. Do not widen scope to drift you notice along the way: report that back instead, so it goes through the normal audit path with evidence attached.
If a finding turns out to be wrong once you open the code, stop. Say so, name what the evidence missed, and close nothing. A wrong fix costs more than a missed one.
When the prompt contradicts the issue, trust neither. Verify.
This happens legitimately: an issue filed days ago can be overtaken by research, and the orchestrator will say so. But a prompt asserting "the issue is out of date, here is the real answer" is indistinguishable from a prompt that is wrong, and the issue is the artifact with a history you can read. Go to the source yourself and settle it.
That is not hypothetical either. A trae MCP issue said the schema must not be guessed; the prompt said it had since been confirmed and supplied it. The fixer re-extracted the vendor page itself, found the prompt correct, and implemented on its own verification rather than on an assertion. Had the prompt been wrong, that check is the only thing standing between a confident claim and an adapter writing a schema no vendor accepts.
Steps
- Read
docs/internal/adding-adapters.mdand theadapter-patternrule before touching an adapter. Read a neighbouring adapter for file shape. - Branch from fresh
origin/main. Name it for the bucket, not the finding:fix/target-audit-<target>for a breaking fix,feat/target-audit-native-surfacesfor a batched additive PR,docs/target-audit-<date>for the docs-only bucket. - Write the failing test first. For a path change that is the target's
golden test under
internal/adapters/<target>/testdata/; for a new surface it iscapability_parity_test.go, which fails as soon as you add the kind tocaps.Supportsand before you emit anything for it. - Fix the adapter. Every finding touches some subset of:
internal/adapters/<target>/: emission pluscaps.Supports- the
importside, when the moved path is one we read back - the adapter package doc comment, which states what the tool reads
docs/user/targets.md: capability matrix row and per-target section.agnostic-ai/skills/target-audit/references/sources.md, when a URL moved
agnostic-ai sync, then commit the regenerated per-target files. A PR that leavessync --checkred will fail CI.make preflightandagnostic-ai sync --checkmust both pass before you push. Never push red.- Add a
[Unreleased]entry toCHANGELOG.mdunderAdded,Changed, orFixed. One line, user-facing effect, no em dashes. - Push and open the PR:
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.
- 7d ago First seen · 92 lines · 22 tokens per session scan A 7000a8e67cdb
adapter-fixer is an agent published in the GitHub repository Chemaclass/agnostic-ai (11 stars, last pushed 3d ago), licensed MIT. It adds 22 tokens to every session and 1,033 once invoked, about $0.0001 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 2026-08-30.
Other agents, from other repositories
tldrcrew-reviewer
Diff/branch/file reviewer. One line per finding, severity-tagged, no praise, no scope creep. Output format path:line: : . . Use for "review this PR", "review my diff", "audit this file". Skips formatting nits unless they change meaning.
oracle
Strategic technical advisor. Use for architecture decisions, complex debugging, code review, simplification, and engineering guidance.
momus
Judge. Ruthless code and plan reviewer. Rejects bad work. Use for final verification loops.
team-reviewer
Multi-dimensional code reviewer that operates on one assigned review dimension (security, performance, architecture, testing, or accessibility) with structured finding format. Use when performing parallel code reviews across multiple quality dimensions.
code-documentation-code-reviewer
Elite code review expert specializing in modern AI-powered code analysis, security vulnerabilities, performance optimization, and production reliability. Masters static analysis tools, security scanning, and configuration review with 2024/2025 best practices. Use PROACTIVELY for code quality assurance.
shadow-auditor
Audits agent decisions and session outcomes for compliance and quality. Assign as a shadow for end-of-session review.