The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local scripts (scripts/validate.sh) and audio utilities (ffmpeg, ffprobe) to validate sequence syntax and analyze rendered audio properties. These commands are executed with structured arguments and follow a strictly read-only policy for user source files.
[EXTERNAL_DOWNLOADS]: The skill is designed to verify the resolution of remote audio resources (ambiance and music) provided via URLs within the sequence files. This is a documented feature used for auditing dependency accessibility.
[REMOTE_CODE_EXECUTION]: The skill includes a fallback in its validation script that uses go run to execute its validator from the local repository source. This is a standard developer-oriented execution path for the tool's own logic and does not involve untrusted remote code.
[PROMPT_INJECTION]: The skill ingests untrusted user-controlled .spsq files, which represents an indirect prompt injection surface when generating reports or handoff prompts for other skills.
Ingestion points: Reads complete .spsq and .spsc files for analysis (SKILL.md).
Boundary markers: Enforces a strict read-only stance and utilizes a specific report structure (references/report-format.md).
Capability inventory: Can execute local validation scripts and media analysis tools, and can generate transition prompts for the create-spsq skill.
Sanitization: No explicit sanitization of file content is described for handoff prompts, though instructions require concrete findings and preservation of sequence identity.

review-spsq