The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local commands like pwd, ls, and find to analyze the project environment and detect existing source code files to provide context for the generated specifications.
[EXTERNAL_DOWNLOADS]: The skill's 'Auto Mode' fetches requirement data from GitHub Issues using the gh issue view command. GitHub is recognized as a well-known and trusted service.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external GitHub Issues, creating a vulnerability surface where malicious content in an issue could influence the agent's output.
Ingestion points: Data is fetched via gh issue view in references/auto-mode.md and used to drive the full specification workflow.
Boundary markers: The skill uses an assumption system (ASM-XXX) and out-of-scope definitions to clarify scope, but it does not include explicit instructions to sanitize or ignore potentially malicious prompts within the ingested issue body.
Capability inventory: The skill can discover files (ls, find), write markdown documents, and is instructed to execute generated verification commands during the test planning phase.
Sanitization: No specific sanitization or filtering logic is applied to the Issue content prior to processing.
[DYNAMIC_EXECUTION]: The skill generates commands for the acceptance test plan (test.md) and is instructed in references/test-plan.md to 'actually run' or dry-run these commands at generation time to verify their validity. This represents the execution of dynamically generated content based on the project's requirements.

spec-generator