The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches a GitLab CI template from a remote GitHub repository belonging to the author organization and references a GitHub Action for pipeline integration. Both resources are hosted on a well-known service and are consistent with the skill's stated purpose.
[COMMAND_EXECUTION]: The skill uses package runners such as npx and pnpm dlx to run the tool's recommendation and migration commands, and it uses standard package managers to install the tool as a development dependency.
[PERSISTENCE]: The fallow agent install command establishes commit and push gates by writing hooks to configuration files like .claude/settings.json and .codex/hooks.json. It also installs agent skills into the local directory structure to maintain quality checks across development sessions.
[INDIRECT_PROMPT_INJECTION]: The skill processes project configuration files to identify existing tools and generate recommendations.
Ingestion points: Files detected during the inspection phase, including package.json, tsconfig.json, and configurations for tools like Knip or jscpd.
Boundary markers: The instructions explicitly state to treat project config as untrusted input and prohibit the addition of remote configuration extensions.
Capability inventory: The skill performs file system writes (writing configs, skills, and hooks), executes package installations, and registers MCP servers.
Sanitization: The skill mandates the use of --dry-run for initial mutating commands and includes logic to ask for user approval before applying subjective configuration changes.
[DYNAMIC_EXECUTION]: The skill dynamically generates and writes agent skills and server registrations to the repository during the agent wiring phase.

fallow-setup