The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text from an interview process and data retrieved from local code mining tools to generate new skill files. This architectural pattern creates an interface where external data could potentially influence the logic of the newly created skills. • Ingestion points: User input from the interview process defined in Step 1 and outputs from the pythia CLI tools in Step 2. • Boundary markers: The generated content is placed into a structured markdown template, though explicit sanitization for user-provided strings is not programmatically enforced within the skill itself. • Capability inventory: The skill facilitates file system writes to create new skill files in .claude/skills/ and executes local CLI utilities. • Sanitization: The skill contains explicit documentation instructing the user to perform a hygiene check for credentials and internal identifiers before sharing generated skills.
[COMMAND_EXECUTION]: The skill directs the agent to utilize CLI tools like pythia and npx to perform code analysis and manage skill deployment. These operations are restricted to the local environment and are standard components of the intended developer workflow.

pythia-skill-author