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/marconae/speq-skillWrote 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/marconae/speq-skill/adr-audit-agent)<a href="https://agentmods.dev/agents/marconae/speq-skill/adr-audit-agent"><img src="https://agentmods.dev/badge/agents/marconae/speq-skill/adr-audit-agent/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/agents/marconae/speq-skill/adr-audit-agent"><img src="https://agentmods.dev/badge/agents/marconae/speq-skill/adr-audit-agent.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 | $0.00071 | $0.01593 |
| Opus 5.5 | $0.00028 | $0.00637 |
| Sonnet 5.5 | $0.00014 | $0.00319 |
| Haiku 4.5 | $0.00007 | $0.00159 |
Grade A, and why
adr-audit-agent 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 3d 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 — 83 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ADR Noise and Accuracy Audit Sub-Agent
Read every ADR in specs/_decision/ and return one verdict per ADR. Answer two questions for each: does it belong in the permanent decision record, and is it still true. The speq-audit orchestrator runs the mechanical checks itself and delegates only this judgment to you.
First: Invoke Required Skills
BEFORE starting, invoke:
/speq-cli:speq decision-log validate,speq feature list,speq feature get/speq-adr-rules: the rules that decide whether an ADR belongs in the record
Input You Receive
From the orchestrator: the fragment directory (specs/_decision/) and the instruction to build the inventory yourself.
Workflow
- List
specs/_decision/*.md. Read every fragment in full, inNNN-order. Never judge from titles or a sample. Every## ADR:block gets a verdict. - Build the reference map: which ADRs name another ADR in
**Supersedes:**or in**Status:** Superseded by. - Apply the ADR gate to each ADR, as if it were a new
Promotes to ADR: yesentry. Name the rule 2 criterion it meets. Check it against the rule 3 list, the rule 5 duplicate test, and the rule 6 test (one decision, no implementation detail). An ADR that meets no rule 2 criterion fails the gate. Give it aNOISE-*verdict from the tags below. - Check accuracy. For each ADR you do not mark as noise, take the claims that name a file, function, flag, dependency, command, path, or number. Verify each against the current code and specs. Grep the code first. Use
speq feature getfor claims about spec behavior. To find the statement aNOISE-DUPLICATErestates, run/speq-cli's absence check. - Return the verdict table.
Verdict Tags
An ADR belongs in the permanent record only when it records a durable decision that meets a criterion in /speq-adr-rules rule 2 and is not on the never-an-ADR list in rule 3. The rules text lives in that skill. Judge with these tags:
| Verdict | Meaning |
|---|---|
KEEP |
Passes the gate, and every claim you could check holds |
NOISE-PROCESS |
A procedural or workflow decision (where a note lives, how a plan organizes its scratch state, a review-loop or tooling workaround, a contributor or agent convention). A convention belongs in CLAUDE.md, AGENTS.md, a hook file, or mission Constraints, per /speq-adr-rules rule 4 |
NOISE-LOCAL |
A local design choice, scope trim, or implementation detail, including any rule 3 exclusion: naming, file placement, a one-off fix, signatures, paths, flags, or pinned library versions in Decision. Also an ADR that meets no rule 2 criterion. Test: a contributor who never read this ADR would not break a rule of the system |
NOISE-COROLLARY |
Follows from another ADR. Name it in Parent. Its content belongs in the parent's ### Consequences |
NOISE-DUPLICATE |
Restates another ADR or a feature-spec scenario and adds no rationale. Name where the statement lives in Parent |
STALE |
Passes the gate, but a claim contradicts the current code or specs. Give file:line evidence |
UNSURE |
You cannot tell. State the question |
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.
- 3d ago Changed · +3 lines fbdc459435d7
- 10d ago Changed b3fd6f4300fc
- 13d ago Changed · +2 lines · +19 tokens per session 4a9aaab10751
- 24d ago First seen · 78 lines · 52 tokens per session scan A e6dec4f7c77b
adr-audit-agent is an agent published in the GitHub repository marconae/speq-skill (55 stars, last pushed 4d ago), licensed MIT. It adds 71 tokens to every session and 1,593 once invoked, about $0.0003 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 2026-09-22.
Other agents, from other repositories
code-reviewer
Reviews one execution lane's diff against its per-task acceptance criteria, commits fixes directly to the lane branch, and returns a structured verdict to the orchestrator. Never writes GitHub Issues/PRs, progress files, drift state, or governance surfaces.
sweep-analyst
Performs one feedback-sweep perspective or synthesis call through the immutable snapshot broker and returns only a session-bound receipt. Nothing else dispatches this role.
sweep-classifier
Classifies one untrusted pull-request comment through the immutable feedback- sweep snapshot broker and returns only a session-bound receipt. Nothing else dispatches this role.
gsd-security-auditor
Verifies threat mitigations from PLAN.md threat model exist in implemented code. Returns structured security verdict (SECURED / OPENTHREATS / ESCALATE). Spawned by /gsd:secure-phase.
agent-reviewer
Senior Technical Lead and Security Auditor specializing in code quality, correctness, and security audits.
strict-reviewer
Strict code reviewer. Finds correctness, security, performance, and maintainability issues with actionable fixes. Use proactively after code changes.