The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes system commands for version control and file archiving. * Evidence: scripts/lib/index.mjs uses execSync for git rev-parse and git diff to track file changes and provenance. * Evidence: scripts/change-archive.mjs uses execSync to run the system zip utility on completed change workspaces. * Note: Command inputs such as change IDs and file paths are sanitized or derived from trusted local state, minimizing injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill parses untrusted data from the local repository to determine next steps. * Ingestion points: manifest.yaml, architecture.md, decisions.md, and plan.md in scripts/change-status.mjs and scripts/manifest-gate.mjs. * Boundary markers: Absent. The skill reads the files directly without wrapping content in safety delimiters. * Capability inventory: Shell command execution via git and zip in scripts/lib/index.mjs and scripts/change-archive.mjs, along with file read/write operations. * Sanitization: The skill uses a custom, minimal YAML parser in scripts/lib/index.mjs to ingest data, providing some protection against complex parser exploits, but lacks semantic sanitization of the values read.

what-now