The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides functionality to import pattern data from an external file (.cco/patterns.digest.json). This creates an ingestion surface where data controlled by third parties (e.g., via a git repository) could influence the agent's internal categorization of files.
Ingestion points: The import <file> command in SKILL.md allows the agent to ingest external JSON content.
Boundary markers: The skill instructions explicitly state that the imported file is audited for absolute paths and traversal segments, though it does not mention specific prompt-level delimiters.
Capability inventory: The skill uses Bash to execute a local Node.js script (patterns-share.js) and possesses Read access to the file system.
Sanitization: The skill logic claims to reject any files containing absolute paths or traversal segments to prevent file system abuse during the import process.

cco-patterns