Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add AleksandarBisevac/claude-plugins/plugin install auditWrote 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/aleksandarbisevac/claude-plugins/run)<a href="https://agentmods.dev/commands/aleksandarbisevac/claude-plugins/run"><img src="https://agentmods.dev/badge/commands/aleksandarbisevac/claude-plugins/run.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 | $0.00032 | $0.01116 |
| Opus 5 | $0.00016 | $0.00558 |
| Sonnet 5 | $0.00006 | $0.00223 |
| Haiku 4.5 | $0.00003 | $0.00112 |
Grade A, and why
run 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 6d 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 — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/audit:run — execute one task
$ARGUMENTS = the task id to run (plus optional --dry-run). Read
${CLAUDE_PLUGIN_ROOT}/reference/orchestrator.md and
${CLAUDE_PLUGIN_ROOT}/reference/manifest-conventions.md first.
If --dry-run is present: follow the orchestrator's Dry-run / preview section — preview
whether the task is ready + what would run, and STOP without mutating.
Otherwise run the full preflight (steps 1–5, including the lock) and emit Progress output as you go.
Execute exactly <taskId>, with status guards:
status == "done"→ refuse: report itscommit/outcome. Offer (AskUserQuestion) an explicit re-open: on confirmation, resetstatus = "pending",attempts = 0, clearcommit,outcome,completedAt,verifiedBy— and iftask.bugIdis set, reopen the linked bug too (itsbugs[]entry back tostatus: "in_progress", clearfixedIn) so a re-opened bugfix task never leaves its bug markedfixedat a stale SHA — then execute. Never silently re-run a done task. A reopened task with anadolink gets the ADO echo (orchestrator.md → "ADO echo"): its card moves back to the pending-state with the commentreopened by /audit:run— the reopen was human-confirmed, so the board move inherits that consent.testEvidenceis deliberately not on that list. The block is a cache of a run that really happened, the evidence ledger still holds that run, andrun-test-gate.py --reconcilere-derives every subject's pointer from that ledger — so a cleared pointer is put straight back by the next reconcile or the next recorded run, and clearing it buys a reader nothing but a disagreement between the plan and the record, which/audit:doctorwill then report. What marks the verdict as stale is the task readingpendingagain beside anatstamp older than the reopen, not a missing block.status == "blocked"→ refuse: report why (exhausted attempts / blockers). Offer a confirmed reset ofattemptsto 0 (back topending), then execute.status == "in_progress"→ warn: likely an interrupted run — point to/audit:resume. Proceed only if the human explicitly confirms re-execution.- Unmet blockers → refuse and list them.
- Otherwise run Execute the task (orchestrator).
Then follow Reporting and release the lock.
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.
- 6d ago First seen · 69 lines · 32 tokens per session scan A 4f968e21dadf
run is a command published in the GitHub repository AleksandarBisevac/claude-plugins (4 stars, last pushed 5d ago), licensed MIT. It adds 32 tokens to every session and 1,116 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 commands, from other repositories
sdlc-implement
Standalone issue implementation with Ralph Loop — implements a single Linear issue or free-form spec using fullstack-engineer + self-correcting retry loop.
sdlc
Master orchestrator for the complete 5-phase SDLC pipeline. Runs all phases sequentially with approval gates and handoffs between each phase. Organizes artifacts into handoffs/, docs/, and projects/ directories.
to-issues
Create one Linear issue per slice from an existing scope.json. Run standalone to (re)create issues without rerunning the full Plan phase. Falls back to docs/issues.md if Linear MCP is unavailable.
sdlc-plan
Phase 1 — Plan. Runs product-manager grill-me interview + slice decomposition, then publishes PRD and creates issues. Invokes exactly 1 agent.
poam-review
Review and manage Plan of Action & Milestones (POA&M).
spec
SDD 前半段：需求釐清（反問+挑戰）→ OpenSpec change 一次落檔（proposal＋delta＋tasks，Draft）→ TPM 拍板鎖定（回一個 OK 即可）。鎖定後 build 全自主執行。.