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 agentmods add commands/wgo-audit/code/auditgit clone --depth 1 https://github.com/wgo-audit/codeWhat 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 | $0.00020 | $0.01448 |
| Opus 5 | $0.00010 | $0.00724 |
| Sonnet 5 | $0.00004 | $0.00290 |
| Haiku 4.5 | $0.00002 | $0.00145 |
Grade A, and why
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 yesterday.
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 — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/audit
Invoke this as wgo:audit <reviewer-id>/wgo:audit all in Codex,
/wgo:audit <reviewer-id>//wgo:audit all in Claude, or
/wgo-audit <reviewer-id>//wgo-audit all in OpenCode.
Load and use the WGO skill. Resolve the newest dated audit root with its
references/common/audit-root.md. Read and write only that root.
One Reviewer
With <reviewer-id>, confirm that reviewer is selected in audit-brief.md,
then:
- Read the brief, the reviewer card, its relevant predecessor handoffs, and only named shared evidence packets and linked artifacts relevant to its question. Resolve the approved portable package locator from the brief to an absolute runtime path, verify its ID and version, and do not persist that runtime path. On a same-root resume, also read its prior open items and decision inventory/register before assigning an OI, ADR, or PDR ID.
- When the brief names automatic GitHub code repository sources, create or refresh
github-history-and-hosted-ci.mdbefore reviewing. It inspects the named repositories' accessible PRs, issues, Projects, Actions, releases, and history without another approval, and writes only an evidence packet. - Inspect available evidence before selecting outputs; register reusable evidence and material access events. Run any other card-listed shared collector when it supports the question and its packet is absent or stale.
- Select the reviewer card's required outputs and any conditional output whose plain-language trigger is met. Do not create routine not-applicable rows.
- If the reviewer returns a qualifying mandate, acceptable-outcome, priority, or authority question, the coordinator asks it and records the answer before dependent work. The reviewer never contacts the auditor. Use a verification when proof of fact, access, or live state is required.
- Create selected artifacts from their templates, labelling uncertainty in the artifact. Record a concise omission and closure route only when a required or triggered output cannot responsibly be created.
- Run the one ephemeral artifact-quality review. Revise the selected outputs once; it never creates an audit artifact or a second approval step.
- Write the reviewer report, including zero or more evidence-linked decision insights when they meet the shared standard, and compact handoff; then update the checklist and material shared open items.
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.
- yesterday First seen · 121 lines · 20 tokens per session scan A b2da42d176a8
audit is a command published in the GitHub repository wgo-audit/code (2 stars, last pushed 9d ago), licensed MIT. It adds 20 tokens to every session and 1,448 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 2026-08-31.
Other commands, from other repositories
krait
Run a complete multi-phase security audit on the target codebase.
krait-poc
Write and run a valid Foundry proof-of-concept that proves (or disproves) a Solidity exploit by asserting the actual harm on a forked chain or against local source.
init-brief
Initialize a briefloop workspace using conversational onboarding.
layout
Audit pipeline: choose how the manifest is stored — sharded (an index plus one file per phase: fewer tokens per phase run, parallel-safe across worktrees) or single-file (one file, one diff, no index). A layout CHOICE, not a version upgrade: both shapes are current, neither goes out of date, and this command moves in…
usage
Audit pipeline: token spend attributed by phase, task, model, author and time — with cache economics, cost-per-task and a usage trend. Read-only (never mutates the manifest).
krait-review
Re-examine findings killed by the Critic's automatic gates. Catches over-filtering without compromising the main report's zero-FP standard.