Arnie016/codex-prompt-templates / skills/auto-skill-safety-external-skill-intake
auto-skill-safety-external-skill-intake
VerifiedSafely inspect and adapt third-party agent/Codex/Claude skill packs before installation. Use when importing external skills, prompt packs, slash commands, Skill Forge candidates, OpenDesign-style catalog stubs, or repos with SKILL.md files, commands, hooks, or installers. Triggers: import skills, install skill pack, evaluate third-party skill, prompt-pack safety, adapt Claude skill to Codex. Skip when authoring a new first-party skill from scratch; use `$skill-pack-factory` and `$automation-skill-forge`.
External Skill Intake Gate
Generated by: Codex Supercharge maintenance automation.
Goal: convert useful external skill ideas into safe Codex artifacts without running untrusted installers or widening local tool access by accident.
Workflow
- Source intake: record repo/path, HEAD or date, license file, disk size,
runtime, and whether any code was executed. Start from README, manifests,
install scripts, hooks, and
SKILL.mdfiles. - Inventory skill surface: list candidate skills, commands, hooks, side-file folders, scripts, external APIs, credentials, and write targets.
- Classify risk:
prompt-only: instructions and references onlyside-files: assets, examples, references, schemas, or templatescommands: shell, package, browser, network, or filesystem actionshooks: conversation capture, prompt injection, or automatic lifecycleinstaller: writes into~/.codex,~/.claude, PATH, services, or deps
- Gate execution: do not run installers, hooks, package scripts, MCP
servers, binaries, or generated commands until reviewed. Use
$mcp-safety-reviewfor broad tool/runtime exposure and$auto-skill-safety-mcp-secret-guardfor configs or examples with secrets. - Adapt, do not blindly copy: prefer distilled workflow text over full upstream bundles unless license, provenance, side files, and triggers are clear. Preserve source, license, date reviewed, and "code executed: no".
- Validate: run
$skill-quality-gateon any new or changed skill before installing it.
Commands
du -sh sources/<repo>
bash plugins/codex-supercharge/scripts/repo_snapshot.sh sources/<repo>
find sources/<repo> -path '*/SKILL.md' -o -path '*/commands/*' -o -path '*/hooks/*'
rg -n "install|hooks|commands|allowed-tools|api[_-]?key|token|secret|postinstall|curl|delete|destructive|browser|oauth|shell" sources/<repo>
python3 "$HOME/.codex/skills/.system/skill-creator/scripts/quick_validate.py" plugins/codex-supercharge/skills/<skill-name>
plugins/codex-supercharge/scripts/skill_audit.sh plugins/codex-supercharge/skills
Output
# External Skill Intake Report
Verdict: accept | adapt | reference-only | reject
Source:
License:
Code executed:
Skill inventory:
Risk classification:
Install/write targets:
Secret or account surface:
Adaptation plan:
Validation:
Skip When
- The skill is already installed and only needs metadata/quality review; use
$skill-quality-gate. - The task is to create original first-party skills from local requirements;
use
$skill-pack-factoryor$automation-skill-forge. - The source is an MCP server or tool runtime rather than a skill/prompt pack;
use
$mcp-safety-review.
References
plugins/codex-supercharge/references/external-skill-intake.mdplugins/codex-supercharge/references/automation-skill-patterns.mdplugins/codex-supercharge/references/skill-patterns.md
Validation
- Every imported skill has provenance, license/trust notes, trigger/skip boundaries, and an explicit execution decision.
- Hook, installer, credential, browser, shell, and account-control surfaces are classified before installation.
- New or changed skills pass
quick_validate.pyandskill_audit.sh.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |