The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied seed data to generate content, which presents a surface for indirect instructions. Ingestion points: Human-provided topic and angle in SKILL.md and references/collaborative-seeding.md. Boundary markers: None specified in the instructions. Capability inventory: Analysis of SKILL.md shows the skill is restricted to text generation and lacks access to shell, network, or file-system tools. Sanitization: No filtering or validation of the input seed is performed.
[METADATA_POISONING]: The skill's metadata contains an author discrepancy. The YAML frontmatter in SKILL.md identifies the author as https://github.com/dmitry, which contradicts the provided author context and the skill's references to https://github.com/Jeffallan.
[EXTERNAL_DOWNLOADS]: The skill recommends a related tool from a remote source. In SKILL.md, the workflow suggests downloading the the-fool skill from the author's repository at https://github.com/Jeffallan/claude-skills. This is documented as a standard functional extension from a vendor-controlled source.

madman