The Agent Skills Directory

[COMMAND_EXECUTION]: The script evaluate-session.sh uses standard system utilities such as grep, sed, jq, and mkdir to manage local files and process session metadata.
[DATA_EXPOSURE]: The skill accesses the local session transcript to determine message counts. This data remains local and is not transmitted to external servers.
[PROMPT_INJECTION]: The skill is designed to process session transcripts which could potentially contain adversarial instructions from past interactions.
Ingestion points: Ingests session data from the file path provided via stdin in the transcript_path field within evaluate-session.sh.
Boundary markers: The shell script does not implement boundary markers or instructions to ignore embedded prompts within the transcript content it processes.
Capability inventory: The bash script performs directory creation and read operations; the associated agent logic has the capability to write new files to the ~/.claude/skills/learned/ directory.
Sanitization: There is no evidence of sanitization or filtering of the transcript content within the provided shell script or configuration.

continuous-learning