The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by reading external markdown files that define agents.
Ingestion points: Reads agent definitions from .claude/agents/*.md and ~/.claude/agents/*.md (SKILL.md).
Boundary markers: The skill does not use delimiters or instructions to ignore embedded commands when parsing these files.
Capability inventory: The skill has access to Bash, Read, and Grep tools (SKILL.md).
Sanitization: No evidence of validation or filtering of external file content is present.
[COMMAND_EXECUTION]: The skill uses ls and grep via Bash to discover and parse files in user and project directories. This behavior is consistent with the stated purpose of agent discovery.
[EXTERNAL_DOWNLOADS]: The install_command in the metadata fetches the skill definition from a GitHub repository using curl. This is a standard installation pattern and uses a well-known service (GitHub).

agent-discovery