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 skills add ktwu01/benchmark-radar --skill maintainer-local-cleangit clone --depth 1 https://github.com/ktwu01/benchmark-radarWrote 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/skills/ktwu01/benchmark-radar/maintainer-local-clean)<a href="https://agentmods.dev/skills/ktwu01/benchmark-radar/maintainer-local-clean"><img src="https://agentmods.dev/badge/skills/ktwu01/benchmark-radar/maintainer-local-clean/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/skills/ktwu01/benchmark-radar/maintainer-local-clean"><img src="https://agentmods.dev/badge/skills/ktwu01/benchmark-radar/maintainer-local-clean.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 | $0.00041 | $0.00540 |
| Opus 5.5 | $0.00016 | $0.00216 |
| Sonnet 5.5 | $0.00008 | $0.00108 |
| Haiku 4.5 | $0.00004 | $0.00054 |
Grade A, and why
maintainer-local-clean 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 today.
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.
What it actually says
Maintainer local clean
This repo squash-merges, so a merged branch is not an ancestor of main. Prove
merges through GitHub, not through git branch --merged.
Steps
- Pull
main:git pull --ff-onlyonmain, thengit fetch --all --prune. - List candidates:
git worktree list,git branch -vv, andgit for-each-ref refs/remotes. - For each branch, worktree, or remote-tracking ref, find its PR number
(branch name, or
gh pr list --state all --head <branch>). Read its state withgh pr view <n> --json state,headRefOid. - Classify:
MERGED: remove the local branch, worktree, and tracking ref.CLOSEDwithout merge: when you remove the ref, tag its tip first withgit tag legacy/<branch> <ref>(local only, never pushed). Do this automatically and do not list it as a finding. If the ref stays, add no tag.OPEN: keep.star-history(any remote): automation branch. Always keep, never report.- Any other branch with no PR: keep, and report it.
- Before removing a worktree, check
git -C <path> status --shortis empty. A worktree whose directory is gone is removed bygit worktree prune. - Present the plan and wait for approval. Remove in this order: worktrees, local branches, remote-tracking refs.
Constraints
- Delete only local state. Never run
git push origin --deleteunless the maintainer names the remote branch. - Keep all tags, including
legacy/*. - Keep ignored build artifacts (
site/*,data/catalog/*,.venv,out/). They regenerate from the build, but a rebuild is slow. - Do not touch
main, the current branch, or a branch checked out in a worktree you are keeping. - If a hook blocks
git branch -dor-dr, give the maintainer the exact command to run. Do not bypass the hook withupdate-ref.
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.
- today First seen · 43 lines · 41 tokens per session scan A 7ff6473ece9c
maintainer-local-clean is a skill published in the GitHub repository ktwu01/benchmark-radar (294 stars, last pushed today), licensed MIT. It adds 41 tokens to every session and 540 once invoked, about $0.0002 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 2026-10-16.
Other skills, from other repositories
merge-drive
Carry a long-running conflicted merge of origin/main into a branch that restructured the tree: read the merge state from the index, resolve by class, and hand it to the next session without losing work. Covers the classes git reports as already-resolved (silent resurrections, directory-rename ports, half-landed…
haven-play
Try a GitHub PR on this machine with haven pr (own worktree, install, stack on a hostname) or haven pr --throwaway (a throwaway sandbox that destroys itself on quit; formerly haven play). Use when someone says 'try a PR locally', 'haven pr', 'haven play', 'haven pr --throwaway', 'run this PR in a sandbox', 'review a…
parallel-pr-review
Use when asked to "review the open PRs", review a batch or stack of pull requests, or run a recurring PR-review pass on a repo — especially with many PRs, stacked branches, conflicts, or security-sensitive changes. Covers grouping, fan-out to review subagents, verdict synthesis, and posting.
code-changes
Change the user's own program, on their machine or through GitHub. Use when a request needs a change to the user's code (instrument tracing, wire the SDK, fix the agent behind a failing scenario, add a run parameter to a connected agent, version a hardcoded prompt) and not when the platform alone can do it (create a…
github
Open a real pull request. Clone a repo, branch, commit, push, and open a PR authored by the LangWatch app on behalf of the requesting user. Use when the user asks to open a PR, fix something in a repo and submit it, send a patch, raise a pull request, or otherwise land a code change on GitHub.
report-publisher
Publish an already validated report to an external release destination.