continuous-learning
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process session transcripts for pattern extraction, which creates a potential surface for indirect prompt injection. Malicious instructions contained within a transcript could potentially influence the agent during the evaluation phase.
- Ingestion points: The
evaluate-session.shscript identifies the session transcript file path from standard input and counts message types within it. - Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious embedded instructions within the ingested transcripts.
- Capability inventory: The skill has the capability to write to the local file system (
mkdir -p, and instruction to save learned skills) and execute local commands via hooks. - Sanitization: The script does not perform sanitization or filtering of the transcript content before signaling the agent to evaluate it.
- [COMMAND_EXECUTION]: The skill instructs the user to configure a shell script hook (
evaluate-session.sh) that executes automatically at the end of every agent session. While this is the intended mechanism for automated learning, it establishes a recurring execution pattern on the host machine.
Audit Metadata