The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to discover and read the frontmatter and content of local SKILL.md files located in .claude/skills and plugins/ directories to generate summaries for AGENTS.md.
Ingestion points: Content is ingested from local skill files found via shell discovery commands.
Boundary markers: The instructions do not define delimiters or specific warnings to ignore embedded instructions within the ingested skill files.
Capability inventory: The skill performs file discovery (find, ls) and file creation/writing (to AGENTS.md).
Sanitization: There are no instructions to validate, filter, or sanitize the metadata or instructions extracted from the local skill files before they are included in the new documentation.
[COMMAND_EXECUTION]: The skill includes shell commands for project environment setup and file discovery.
The instructions provide commands for creating symlinks (ln -s AGENTS.md CLAUDE.md) and searching the file system (find .claude/skills -name "SKILL.md"). These are standard administrative operations within the context of managing project documentation.

agents-md