The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: A remote code execution pattern (curl ... | sh) was identified in scripts/test_recipe_scanner.py. Detailed analysis shows this is a negative test case for a security validator (scripts/recipe_scanner.py) to verify its ability to detect and block malicious command patterns. It is not part of the skill's executable logic.
[COMMAND_EXECUTION]: Documentation files, such as references/claude-code-routines.md, contain example curl commands for API interaction. These are provided as educational implementation references and do not represent a command execution vulnerability in the skill itself.
[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for processing data from external triggers like webhooks and queues (found in references/webhook-and-queue-triggers.md), which are ingestion points for untrusted data. It mitigates this risk by documenting boundary markers like schema validation and idempotency keys. The skill's capability inventory includes file system access and subagent management, while sanitization is addressed through the recipe_scanner.py utility which validates task blueprints before they are loaded into the runtime.
[SAFE]: The skill's primary logic and scripts adhere to security best practices. The recipe_scanner.py utility uses yaml.safe_load and implements a security gate to flag high-risk extensions and remote code execution patterns.

ai-coding-agents-state