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 skills/buildingopen/bouncer/skillnpx skills add buildingopen/bouncer --skill skillgit clone --depth 1 https://github.com/buildingopen/bouncerWhat 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 | $0.00121 | $0.00847 |
| Opus 5 | $0.00060 | $0.00424 |
| Sonnet 5 | $0.00024 | $0.00169 |
| Haiku 4.5 | $0.00012 | $0.00085 |
Grade A, and why
bouncer 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 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.
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 — 92 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bouncer - On-Demand Quality Audit
Works for any task type: code, creative work, videos, advice, research, operations.
Mode Selection
- Quick (default):
/bounceror "score" or "audit" - Deep:
/bouncer deepor "deep audit" or "verify everything"
Quick Audit
Step 1: Gather context (adapt to the task type)
For code changes: Run git diff:
git diff --stat --no-color
git diff --no-color
git diff --cached --stat --no-color
git diff --cached --no-color
For non-code work (videos, stories, advice, research): Skip git diff. The assistant_text summary is the primary input.
Step 2: Read CLAUDE.md from the current working directory (if it exists).
Step 3: Build a JSON object and pipe it to the bouncer script:
assistant_text: THE MOST IMPORTANT FIELD. A thorough summary of what you did this session. Be specific:- For code: files changed, bugs fixed, tests written, commands run
- For creative: what you created, the goal, the audience, key decisions made
- For videos: the script/story, visual choices, duration, format, iterations
- For advice: the question asked, your recommendation, reasoning, alternatives considered
- For research: what you found, sources checked, conclusions drawn
- CRITICAL: If the staged diff includes changes from prior sessions, note this.
diff_stat: Git diff stat (empty string if no code changes or not applicable)diff_text: Git diff (empty string if not applicable)context: The CLAUDE.md contents (empty string if no CLAUDE.md)
Step 4: Run:
echo '<the JSON object>' | python3 ~/.claude/skills/bouncer/scripts/bouncer-check.py
Step 5: Present results exactly as printed.
Deep Audit
For code-heavy tasks where you want Gemini to independently verify claims.
Step 1: Gather context (same as quick audit).
Step 2: Read CLAUDE.md from the current working directory (if it exists).
Step 3: Build a JSON object with:
assistant_text: Same thorough summary as quick auditdiff_text: Git diff (empty string if not applicable)context: The CLAUDE.md contents (empty string if no CLAUDE.md)cwd: The current working directory (absolute path)
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 92 lines · 121 tokens per session scan A e3cdf567e1d0
bouncer is a skill published in the GitHub repository buildingopen/bouncer (5 stars, last pushed 4mo ago), licensed MIT. It adds 121 tokens to every session and 847 once invoked, about $0.0006 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
cc-hooks
Multi-agent hook system: author hooks once as rulesync-canonical hooks.json (HookDefinitionSchema), deploy via superskill install / superskill hook emit to Claude Code, Codex, OpenCode, Pi (pi-hooks), and Gemini CLI. Covers command, prompt, and http hooks, cross-platform event mapping, and safety invariants.
agent-integration
Run all three agent integration phases sequentially: research, write-tests, and implement using E2E-first TDD (unit tests written last). For individual phases, use /agent-integration:research, /agent-integration:write-tests, or /agent-integration:implement. Use when the user says "integrate agent", "add agent…
development
开发语言能力索引。Python、Go、Rust、TypeScript、Java、C++、Shell。当用户提到编程、开发、代码、语言时路由到此。.
devops
DevOps 能力索引。Git、测试、DevSecOps、数据库。当用户提到 DevOps、CI/CD、Git、测试时路由到此。.
model-update
新增或更新 CCX 项目中的模型注册技能。适用于 Claude 与非 Claude 模型，覆盖能力表、路由白名单、基准映射、代码生成与验证。.
store-update
在 CCX Desktop 发布后下载 Store MSIX 并生成发布公告。用户提到 Store 上架、MSIX、从 GitHub Release 下载 store.msix、发布后同步 Windows Store、从 release 填写商店更新内容时必须使用此技能。该技能会下载最新 GitHub Release 的 amd64/arm64 MSIX，校验 sha256，从 Release body 生成 Store listing releaseNotes 预览，并输出手动上传指引。.