The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script reduce-transcript.py located within the plugin's directory. This script is used to pre-process the transcript before analysis.
Evidence: python3 "${CLAUDE_PLUGIN_ROOT}/scripts/reduce-transcript.py" "<transcript_path>" "/tmp/session-review/reduced.txt" in SKILL.md.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests session transcripts which may contain untrusted data from previous agent interactions (e.g., content from web pages or files).
Ingestion points: The skill reads session transcript files (.jsonl) provided by the user or the environment (SKILL.md, Step 1).
Boundary markers: There are no explicit delimiters or instructions to the subagent to ignore embedded commands within the transcript data (SKILL.md, Step 3).
Capability inventory: The skill possesses the ability to execute shell commands (via python3) and invoke subagents (SKILL.md, Step 2 & 3).
Sanitization: No sanitization or filtering of the transcript content is performed before it is passed to the reviewer agent.

review-session