The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches installation scripts for the Claude CLI from Anthropic's official domain and the skill-validator tool from its GitHub repository.
[REMOTE_CODE_EXECUTION]: Performs piped execution of the Claude CLI installation script via bash. This is an expected installation method from a trusted vendor.
[COMMAND_EXECUTION]: Executes the skill-validator and claude command-line tools to process skill files. It also uses standard utilities like cat, ls, and mkdir for environment setup and state persistence.
[PRIVILEGE_ESCALATION]: Includes instructions for setting executable permissions using chmod +x during manual tool installation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided skill files for analysis, which presents a surface for instructions embedded in the analyzed data to influence the agent. • Ingestion points: Reads SKILL.md and referenced files at a path specified by the user in Step 4 and Step 5. • Boundary markers: No specific delimiters or 'ignore' instructions are used to isolate the analyzed content. • Capability inventory: Capable of executing shell commands (skill-validator, claude) and performing file system operations. • Sanitization: Ingested content is evaluated for quality without explicit sanitization or filtering of embedded instructions.

review-skill