pp-learn-loop-example

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill directs users and agents to install software via 'npx' from the '@mvanhorn/printing-press-library' package and provides links to download pre-built binaries from the author's GitHub repository. These downloads are within the author's own infrastructure and are intended for deployment and updates.
  • [COMMAND_EXECUTION]: The skill's primary function is to execute the 'learn-loop-example-pp-cli' binary to perform API operations and manage a local SQLite database for storing learned patterns and playbooks.
  • [DATA_EXFILTRATION]: The CLI supports a documented '--deliver webhook:' flag that allows command output to be posted to an external URL. While designed for workflow automation, this represents a capability that could be abused to send sensitive data to an untrusted endpoint if the agent is influenced by malicious input.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes natural language query strings through its 'recall' and 'teach' commands, creating a surface for indirect prompt injection.
  • Ingestion points: User queries enter the system via the 'recall' and 'teach' command arguments in SKILL.md.
  • Boundary markers: No explicit boundary markers are used when interpolating these queries into shell commands.
  • Capability inventory: The skill can perform network requests to an API, write to a local SQLite database, and send data to external webhooks.
  • Sanitization: The skill includes a 'ScanPII' function that checks for email addresses and phone numbers in queries, issuing warnings on stderr but not blocking execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 11:53 AM