The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill executes local scripts (scripts/validate.sh) and uses go run to compile and execute source code during the validation process. These tools are provided by the vendor to audit the .spsq file format.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided files (.spsq and .spsc) and reflects their content into reports and 'handoff' prompts for other agent skills. This creates an attack surface where instructions embedded in the sequence data could potentially influence downstream tasks like file creation or behavior explanation.
Ingestion points: SKILL.md (Resolving the review scope section), references/review-checklist.md.
Boundary markers: None identified; the skill does not explicitly sanitize or wrap audited content to prevent instruction following by the next agent.
Capability inventory: Subprocess execution via bash, go run, ffmpeg, and ffprobe.
Sanitization: None identified for the handoff prompt generation.

review-spsq