The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Downloads and executes the Claude Code installer from the official claude.ai domain using a pipe to shell.
[EXTERNAL_DOWNLOADS]: Installs the skill-validator tool from an external GitHub repository (agent-ecosystem/skill-validator) using go install or Homebrew.
[COMMAND_EXECUTION]: Invokes several system and third-party CLI tools including skill-validator, claude, brew, and go to perform validation, scoring, and installation tasks.
[PRIVILEGE_ESCALATION]: Includes manual installation instructions that involve moving binaries to /usr/local/bin and applying executable permissions with chmod +x.
[PERSISTENCE]: Creates and updates a state configuration file at ~/.config/skill-validator/review-state.yaml to persist prerequisite check results between sessions.
[INDIRECT_PROMPT_INJECTION]: The skill reads and evaluates user-provided SKILL.md and reference files, creating a potential vector for instructions embedded in those files to affect the agent's behavior or the scoring outcome.
Ingestion points: Reads file contents from user-specified paths in Step 4 of SKILL.md.
Boundary markers: No delimiters or isolation instructions are used when processing the external content.
Capability inventory: The ingested content is processed by skill-validator and the claude CLI tool.
Sanitization: The skill performs no validation or sanitization of the contents before processing.

review-skill