The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from various sources within the repository and the user's home directory.
Ingestion points: The skill reads local project files including PRODUCT.md, DESIGN.md, product-facts.md, and existing briefs located at docs/point-of-view/<slug>.md or ~/.claude/point-of-view/<slug>.md (SKILL.md).
Capability inventory: The skill is capable of writing and updating markdown files in the local repository and the user's .claude configuration directory (SKILL.md).
Boundary markers: The skill does not explicitly use delimiters or boundary markers when reading existing documentation to prevent the interpretation of data as instructions.
Sanitization: A security control is present where the agent must show the generated brief to the user for confirmation or correction before saving it to the filesystem (SKILL.md).

craftsmans-wisdom