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/ai-sdlc-framework/ai-sdlc/executegit clone --depth 1 https://github.com/ai-sdlc-framework/ai-sdlcWhat 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 | $0.00051 | $0.29813 |
| Opus 5 | $0.00026 | $0.14907 |
| Sonnet 5 | $0.00010 | $0.05963 |
| Haiku 4.5 | $0.00005 | $0.02981 |
Grade C, and why
execute scanned grade C with 1 finding 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 2d 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.
Tells the agent never to refusehighAnti-refusal
Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.
Heuristic 3 is intentionally conservative — it only fires when BOTH conditions hold. A local session without a signing key fails later at Step 10; the heuristic targets the case where an operator explicitly uses a manage How it starts
The opening of the file, as written. The whole thing — 1,760 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Execute work item $ARGUMENTS end-to-end. $ARGUMENTS is either a backlog task ID (e.g. AISDLC-393, INGEST-42) or a GitHub issue (612, #612, or explicit gh:612) — see Argument forms below. The Step 0-15 pipeline below runs inline in the main Claude Code session — worktree creation, developer subagent fan-out, 3 parallel reviewer subagents, attestation signing, PR open.
Argument forms (AISDLC-393)
/ai-sdlc execute accepts three argument forms, detected via three regexes evaluated in this order:
| Form | Regex | Routes to | Example |
|---|---|---|---|
| Explicit GitHub issue | ^gh:\d+$ |
GH-issue path (subscription billing) | /ai-sdlc execute gh:612 |
| Prefixed backlog task ID | ^[A-Za-z][A-Za-z0-9]*-\d+(\.\d+)*$ |
backlog-task path (existing Step 0-15 flow) | /ai-sdlc execute AISDLC-393 |
Bare numeric / #-prefixed |
^#?\d+$ |
GH-issue path (subscription billing) | /ai-sdlc execute 612 / /ai-sdlc execute #612 |
The gh: form has highest precedence so an operator can be unambiguous when they need to be (e.g. when a task-id-like string would otherwise be matched). The regex shapes are the contract — the JS reference implementation lives in dogfood/src/dispatch-execute-arg.ts (parseExecuteArg) with hermetic coverage in dispatch-execute-arg.test.ts; the shell pipeline in Step 1 mirrors the same shapes for the slash-command body.
Both paths use the subscription SubagentSpawner — the difference is the source of truth for the work item:
- Backlog-task path: backlog file in
backlog/tasks/<id> -*.mdcarries the title, ACs, references, andpermittedExternalPaths. Task lifecycle is closed by moving the file tobacklog/completed/in the PR. - GH-issue path: GitHub issue is the source of truth (NO backlog task file is created). Issue title + body + labels feed the developer prompt; PR body uses
Closes #Nso the issue auto-closes on merge.
If $ARGUMENTS matches none of the three regexes, the command exits 1 with a clear error listing the accepted forms (AC-6).
Preserved unchanged (AC-7): the existing watcher pnpm --filter @ai-sdlc/dogfood watch --issue <id> continues to work for API-key billing / unattended / CI use cases. It accepts the same forms as the slash command via the same parseExecuteArg parser.
AISDLC-218 — 1 CI run per PR. Prior to this change, opening the PR before reviewers completed triggered CI run #1 (failing verify-attestation), then the attestation chore commit triggered CI run #2. The fix: the developer opens the PR as a draft (
gh pr create --draft). Reviewers run + attestation signs while still draft. Step 13 callsgh pr readyto flip draft→ready_for_review, which triggers CI exactly once on the fully-signed, reviewer-approved state. ~50% CI-minute reduction per 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.
- 2d ago First seen · 1,760 lines · 51 tokens per session scan C ca3436dbc6c2
execute is a command published in the GitHub repository ai-sdlc-framework/ai-sdlc (92 stars, last pushed 8d ago), licensed Apache-2.0. It adds 51 tokens to every session and 29,813 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 1 finding (tells the agent never to refuse). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
discover
Discover runtime AI components and static AI dependencies with AIR Blackbox.
evidence
Export a signed compliance evidence package as JSON or PDF for regulatory handoff.
replay
Replay recorded AI traces to detect behavioral drift between model versions.
comply
Run a full EU AI Act compliance scan on your Python AI project.
remediate
Close the debt the deterministic lanes cannot version-bump away: the grandfathered broken build, advisories with no direct upgrade path, the lint backlog, missing tests, missing docs. Each run plans that debt as finite work orders and works them in two tiers: deterministic recipes execute first at $0 (lockfile resync…
metrics
Read-only 3-metric governance glance — override rate, small-lane rate, sprint low-confidence ratio — each with a trend arrow vs. the prior 20-commit window.