The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local developer-provided test script (dev/scripts/audit-tests.sh) as its primary step to verify framework integrity.
[COMMAND_EXECUTION]: The skill invokes the node runtime to perform version verification on package.json and syntax validation on the installer file bin/install.js.
[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it processes the contents of various local files including skills, agents, and templates.
Ingestion points: Multiple source files located in gsp/skills/, gsp/agents/, gsp/templates/, as well as CHANGELOG.md, VERSION, and package.json.
Boundary markers: Absent; the content is read directly into the agent's context for analysis.
Capability inventory: Access to Bash for command execution, Agent for spawning sub-agents, and WebFetch for external requests.
Sanitization: None; the skill treats the content of the audited files as data to be grepped and verified.

gspdev-audit