The Agent Skills Directory

[COMMAND_EXECUTION]: The skill dynamically detects and executes shell commands for testing, building, and coverage analysis based on values found in external configuration files.
Evidence: Detection logic in Step 5b and 5c extracts test_command and build_command from openspec/config.yaml, package.json, pyproject.toml, and Makefile for immediate execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple files within the repository to drive its validation logic and generate behavioral compliance matrices.
Ingestion points: Reads implementation tasks from tasks.md, requirements from specs/, design decisions from design.md, and execution parameters from openspec/config.yaml.
Boundary markers: No explicit boundary markers or instruction-separation protocols are defined for handling the content of these files.
Capability inventory: The agent has permission to read files, write verification reports to the filesystem (openspec/changes/), and execute detected shell commands.
Sanitization: No sanitization or validation of the content within specs/ or config.yaml is performed before it is used to influence agent behavior or command strings.

sdd-verify