The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill documentation instructs users to execute npx agentic-flow. This command downloads and runs code from the npm registry. Neither the package agentic-flow nor its author 'agentic-flow' are on the list of trusted organizations, posing a risk of executing unverified external code.
[COMMAND_EXECUTION] (MEDIUM): The use of npx combined with parameters like workers agents ultralearn triggers command execution. Since the underlying package is from an untrusted source, these commands could perform unauthorized actions on the host system.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill utilizes a self-learning feedback loop and memory coordination across multiple agents.
Ingestion points: Performance metrics (quality scores, latency) and memory keys ({trigger}/{topic}/{phase}) in SKILL.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded content were found in the provided documentation.
Capability inventory: The skill dispatches tasks to various specialized agents (researcher, coder, security-analyst) based on ingested performance data.
Sanitization: No evidence of sanitization or validation of the ingested feedback or memory content before it influences agent selection logic was found.

worker-integration