The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to inspect the local repository environment, including ls, find, and cat to read configuration files. It also performs filesystem management using ln -s, mv, and readlink to handle workflow symlinks.
[EXTERNAL_DOWNLOADS]: Fetches and installs agent skills using the skills CLI tool from a well-known service. The execution npx --yes skills@1.5.23 add Fandhe-AI/agent-cli-skills uses a specific pinned version (1.5.23) to mitigate supply chain risks. The downloaded skills are associated with the same author as this skill.
[INDIRECT_PROMPT_INJECTION]: Ingests untrusted data from the local repository to influence the configuration of the Claude Code environment.
Ingestion points: Processes the output of ls, find, and the contents of package.json, Cargo.toml, and pyproject.toml in SKILL.md to detect frameworks and languages.
Boundary markers: Absent; the skill does not use specific delimiters or instructions to ignore embedded commands within repository files.
Capability inventory: Possesses capabilities for command execution (npx, gh), file writing (.claude/ directory contents), and symlink manipulation in SKILL.md.
Sanitization: Repository data is processed directly without explicit sanitization or validation of the file contents.

init-claude