The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from analytical reports that could potentially contain instructions intended to influence the agent's code generation logic.
Ingestion points: Files such as report_card.json, run_metrics.json, and optimization-*.md are read from ~/.vibehq/analytics/ to determine necessary code changes.
Capability inventory: The skill has the ability to modify source code in src/ and bin/, and execute shell commands for building the project.
Boundary markers: The instructions do not specify any delimiters or warnings to treat the content of the analytics files as non-executable data.
Sanitization: No sanitization process is described for the fix_actions or improvement_suggestions found in the ingested reports.
[COMMAND_EXECUTION]: The skill uses local shell commands to facilitate the development lifecycle.
It executes npx tsup to compile the TypeScript project.
It leverages project-specific CLI tools like vibehq-analyze and vibehq start for data collection and testing.
[DYNAMIC_EXECUTION]: The skill is designed to perform autonomous engineering by writing and refactoring TypeScript files in the src/ directory, which are then compiled into executable code. This dynamic code modification is the primary function of the skill.

optimize-protocol