The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and process arbitrary content from local folders or remote GitHub repositories to generate documentation. If the source material contains malicious instructions hidden in code comments or markdown, the agent might attempt to follow them during the 'incremental authoring loop'.
Ingestion points: The scripts/inventory.py script and the workflow step involving reading file contents.
Boundary markers: The workflow lacks specific boundary markers or instructions to ignore embedded commands when reading files.
Capability inventory: The agent can execute shell commands via scripts/scaffold_docs.py, write files to the system, and trigger deployments.
Sanitization: No sanitization of the input content is performed before processing.
[EXTERNAL_DOWNLOADS]: The skill performs several external downloads during its normal operation:
Clones remote repositories using git clone during Step 2 of the workflow.
Uses npx to download and execute the latest version of create-fumadocs-app from the npm registry.
[COMMAND_EXECUTION]: The script scripts/scaffold_docs.py utilizes subprocess.run() to execute shell commands, including npx for project scaffolding and dependency installation. This is a functional requirement for the skill's primary purpose.

lovstudio-repo2docs