The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation encourages users to install the tool by downloading a shell script from a remote GitHub repository at https://raw.githubusercontent.com/codesoda/make-prd/main/install.sh.
[REMOTE_CODE_EXECUTION]: The README and the installation script facilitate a pattern where remote code is executed directly via a pipe to bash. This specific execution method is identified by automated security tools as a significant risk vector.
[COMMAND_EXECUTION]: The install.sh script executes several commands to configure the local environment, including git clone to retrieve the source and ln -s to install the skill into application-specific directories like ~/.claude/skills.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted user descriptions to generate PRD content. Evidence: 1. Ingestion points: The skill accepts a feature description in SKILL.md. 2. Boundary markers: None are present to distinguish user input from instructions. 3. Capability inventory: The skill can write files to the tasks/ directory. 4. Sanitization: There is no validation or filtering of user-provided content before it is processed by the agent.

prd