Borrowing it
Nothing to install: this file belongs to xetenet/xete-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/xetenet/xete-mcp/main/.claude/skills/gate-backtest/SKILL.mdgit clone --depth 1 https://github.com/xetenet/xete-mcpWrote 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/xetenet/xete-mcp/gate-backtest)<a href="https://agentmods.dev/skills/xetenet/xete-mcp/gate-backtest"><img src="https://agentmods.dev/badge/skills/xetenet/xete-mcp/gate-backtest/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/xetenet/xete-mcp/gate-backtest"><img src="https://agentmods.dev/badge/skills/xetenet/xete-mcp/gate-backtest.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 | $0.00087 | $0.00870 |
| Opus 5 | $0.00044 | $0.00435 |
| Sonnet 5 | $0.00017 | $0.00174 |
| Haiku 4.5 | $0.00009 | $0.00087 |
Grade A, and why
gate-backtest 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 10d 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 — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Gate Backtest
The repo's own history is a labeled dataset: every fix commit marks an earlier commit that introduced a real defect. This skill replays that history against the gates and converts each real defect into a permanent benchmark.
Safety invariant
All work happens in the throwaway clone that scripts/backtest.sh creates — remotes stripped, nothing writable pointing at the real repo. Never run analysis inside an active working copy. Outputs (benchmarks, pattern proposals) land only in the xete-agent-skills repo via PR.
Step 1: Mechanical replay
scripts/backtest.sh <path-or-url-to-repo>
Produces: recall (fraction of real defect-introducing commits the gate would have caught), friction (gated commits with no known defect), missed culprits with files, and suggested pattern additions.
Step 2: Qualitative pass on every culprit (caught AND missed)
For each culprit commit in the report, answer three questions by reading the actual diff of the culprit and its fix:
- Which gate would have caught it? Map the defect to a specific line of the solana-security-hardening checklist, or to a doubt class in doubt-driven-review. If NO existing gate item maps to it, that is a checklist gap — the fix becomes a new checklist line.
- Was it catchable at commit time? Some defects are only visible with later context (spec changed, dependency behavior). Mark honestly:
catchableorhindsight-only. Hindsight-only cases must not inflate the gate's claimed value. - What is the doubt prompt? Write the one question a fresh-context reviewer would have needed to ask to find it. Concrete, not generic: "what happens when the same commitment is submitted to two tabs?" not "is this secure?"
Step 3: Author benchmark cases
One file per real defect: benchmarks/BM-<short-name>.md
# BM: <defect in one line>
Source: <repo> culprit <hash> fixed by <hash>
Paths: <files/dirs involved>
Class: <overflow | replay | validation-gap | key-leak | logic | ...>
Catchable at commit time: yes | hindsight-only
Gate mapping: <checklist item or doubt class; NEW if it was a gap>
Doubt prompt: <the question that finds this bug>
Real solution: <what the actual fix did, 1-2 lines>
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.
- 10d ago First seen · 59 lines · 87 tokens per session scan A 34413bd8cf72
gate-backtest is a skill published in the GitHub repository xetenet/xete-mcp (0 stars, last pushed 23d ago), licensed MIT. It adds 87 tokens to every session and 870 once invoked, about $0.0004 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 skills, from other repositories
gitlab-merge-request-review
Review and merge GitLab merge requests via the gitlab-api MCP server — by default the MRs assigned to you or where you are a reviewer, optionally scoped to a project, group, or the whole instance. Read an MR, gate it on a green pipeline (pipeline → jobs → job log), then approve, accept/merge, or set…
review-changes
Review current staged, unstaged, and untracked Git changes for correctness and regression risk, grounded in changed paths, diffs, focused file reads, and nearby test discovery. Use when the user asks to review local changes or a working-tree diff; do not use for broad repository audits.
pr-analysis
Analyse a GitHub pull request by fetching its metadata, diff, linked issues and CI status.
address-pr-review-locally
Evaluate and address a GitHub pull request review locally. Use when a PR review contains inline comments, suggestions, or requested changes that need to be implemented in the local workspace. Covers fetching review threads, triaging by severity, applying code fixes, running tests, and optionally replying to or…
commit-suggest
Conventional Commits message generator from current staged + unstaged changes. Read-only — runs only allowlisted git commands, never commits, never pushes, never adds AI co-author trailers. Output is plain text the user pastes. Triggers: "suggest a commit", "commit message", "commit msg", "write a commit", "gen…
enable-journaling
Opt-in setup for auto-journaling capture plumbing. Copies bundled PostToolUse hook scripts into .squad/hooks/ and prints the .claude/settings.json snippet to wire them up. Capture-only — squad behaviour does NOT change until a follow-up release adds distillation + retrieval. Triggers: "enable journaling", "turn on…