The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from VERIFY.md (both project-level and change-level). This data is used to drive agent behavior, specifically determining which shell commands to execute.\n
Ingestion points: openspec/VERIFY.md and openspec/changes/<name>/VERIFY.md (Step 8).\n
Boundary markers: None present. The skill directly parses and executes the content of fenced code blocks.\n
Capability inventory: The skill utilizes the bash tool to execute arbitrary shell commands captured from the files.\n
Sanitization: No sanitization or validation of the commands extracted from VERIFY.md is performed before execution.\n- [DYNAMIC_EXECUTION]: The skill performs dynamic execution by parsing markdown files to extract shell commands and then executing them at runtime via the bash tool.\n
Evidence: Step 8 describes the process of reading VERIFY.md, extracting code blocks under repo-specific headings, and running them using bash.\n- [COMMAND_EXECUTION]: The skill makes extensive use of the bash tool and the openspec CLI to manage changes and perform verification steps.\n
Evidence: Step 1 uses openspec list, Step 2 uses openspec status, Step 3 uses openspec instructions apply, and Step 8 executes arbitrary commands from VERIFY.md (e.g., npm run lint, cargo test).

openspec-verify-change