outsource — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited outsource (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
outsource drives the `convertible` CLI so a Claude agent can hand a scoped task to a different engine (default: a local vLLM Qwen3.6-27B on :8001). Convertible's model is not assumed to be stronger than you — its value is diversity. A second, independent mind catches things the author's mind glides past, which is why review is the headline verb.
This skill is the operator: a portable wrapper that resolves the CLI and turns each verb into a convertible drive, then prints the drive's result summary.
The entry point is scripts/outsource.sh. Invoke it from the repo you want convertible to work on:
bash .claude/skills/outsource/scripts/outsource.sh <verb> "<text>" [options]It resolves the CLI portably — an installed convertible on PATH (the normal case), falling back to uv run convertible when inside the convertible checkout, else an install hint.
| Verb | What it does | Side effects |
|---|---|---|
explore "<question or area>" | Read-only investigation of the repo; the model reads and reports findings. | None — runs in a throwaway worktree at HEAD. |
review "<what to focus on>" [--base main] | A diverse second opinion on the committed diff (<base>...HEAD). | None — throwaway worktree; reviews committed changes only. |
write "<task>" [--pr] | Implement a change. Commits to a drive branch by default; --pr pushes + opens a PR. | In-place: a convertible/<id> drive branch (or a PR). |
| Option | Meaning |
|---|---|
--repo PATH | Target repo (default: .). |
--base BRANCH | Base for the review diff (default: main). |
--engine NAME | Engine wheel (default: $CONVERTIBLE_ENGINE or vllm-openai). |
--model NAME | Model (default: $CONVERTIBLE_MODEL or mmangkad/Qwen3.6-27B-NVFP4). |
--base-url URL | OpenAI base URL (default: $CONVERTIBLE_BASE_URL or http://localhost:8001/v1). |
--max-steps N | Loop step budget (default: 20). |
--allow-dirty | (write) allow running on a dirty tree. |
--pr | (write) push + open a PR instead of a local drive branch. |
The result printed to stdout is the drive's TaskResult.summary (plus changed_files / drive branch for write), parsed from convertible drive --json. Per-step progress streams to stderr while it runs.
want a candid, independent pass over the committed diff before you trust it. Treat the output as a second opinion to weigh, not a verdict.
X work here?") without anchoring on your own assumptions.
result lands on a drive branch you can inspect, merge, or discard.
git worktreeat HEAD, so a stray write can't reach your working tree or branch; the prompts also tell the model not to modify anything. Don't route a change-making task through them — use write.
--allow-dirty. This guardsthe dirty-tree hazard: convertible drive --no-pr commits uncommitted edits onto the drive branch and leaves you there. Commit or stash first.
smaller/different model; weigh its findings, verify its claims, and own the decision yourself.
sandbox — the loop always exposes write_file/run_command, so the model can still run arbitrary read-only commands.
review covers committed changes only (<base>...HEAD). To reviewuncommitted work, commit it first.
fleet (different model per verb) is separate infrastructure.
This is a first-party convertible skill — agentculture/convertible is its origin. guildmaster re-broadcasts it to the mesh (the same inbound pattern as the devague-origin workflow skills), tracking it in docs/skill-sources.md. The cite, don't import policy holds: downstream repos copy it, they don't symlink or depend on it.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.