The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands and vendor-specific CLI tools using uv run to verify authentication and retrieve pipeline status information.
[DATA_EXFILTRATION]: Accesses the local file system at ~/.local/share/mega-code/ to read configuration paths and metadata about pending items. This access is restricted to the skill's own operational data.
[PROMPT_INJECTION]: The skill ingests untrusted data by reading and displaying descriptions from markdown files located in pending-skills and pending-strategies directories.
Ingestion points: Reads description: fields from SKILL.md and titles from various .md files in ~/.local/share/mega-code/data/.
Boundary markers: None identified.
Capability inventory: Uses Bash and Read permissions to list and parse file contents.
Sanitization: Employs head -c 60 to truncate the displayed metadata, which provides a minor degree of mitigation against large-scale injection payloads.

mega-code-status