The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external documentation, source code, and project history to synthesize new skills. This creates an attack surface for indirect prompt injection, where malicious instructions in external sources could influence the agent's behavior or the content of generated artifacts.
Ingestion points: The synthesis and source discovery paths in SKILL.md, references/synthesis-path.md, and references/source-discovery.md describe collecting source material from library documentation, API references, and PR/issue history.
Boundary markers: The current instructions do not explicitly require the agent to use specific delimiters or protective instructions when incorporating external content into new skills.
Capability inventory: The skill has the capability to write files and execute a local Python validator script.
Sanitization: No explicit requirement for sanitizing or filtering synthesized content from external sources is provided.
[EXTERNAL_DOWNLOADS]: The maintenance evaluation workflow in EVAL.md and references/skill-evals.md utilizes the AXIS evaluation framework from Netlify. This represents a dependency on a well-known and trusted service provider.
[COMMAND_EXECUTION]: The skill uses uv run to execute a local structural validator script (scripts/quick_validate.py). The script is designed to perform safe structural checks on YAML frontmatter and file references without performing network operations or using dangerous dynamic code execution.

skill-writer