The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The review_skill_improvement prompt in prompts.yaml is designed to ingest untrusted data from task summaries and execution statistics for the purpose of suggesting skill improvements.\n
Ingestion points: prompts.yaml (task_summary, stats_json, validation_summary).\n
Boundary markers: The prompt includes explicit instructions to treat all supplied text as untrusted evidence and to not follow embedded instructions.\n
Capability inventory: The skill has access to Bash and filesystem Write tools.\n
Sanitization: Relies on instructional guardrails within the prompt.\n- [COMMAND_EXECUTION]: SKILL.md documents ecosystem commands such as dcc-mcp-cli for monitoring health, discovery, and performing updates.\n- [REMOTE_CODE_EXECUTION]: The installation instructions in SKILL.md reference using npx and openclaw to download and install components from the ecosystem's registry.\n- [DYNAMIC_EXECUTION]: The scripts/validate_skill_dir.py script reads and analyzes the Abstract Syntax Tree (AST) of external Python scripts to check for best-practice import violations.

dcc-mcp-skills-creator