delivery-loop-retro

Pass

Audited by Gen Agent Trust Hub on Jul 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a bundled Python script (scripts/collect_trace.py) to automate the collection of performance metrics from local session logs.
  • [DATA_EXFILTRATION]: The skill accesses sensitive session history data located in ~/.claude/projects and agent memory in the project's .claude/agent-memory directory.
  • This access is fundamental to the skill's purpose of generating retrospectives. No network operations or attempts to exfiltrate this data were identified.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from session transcripts and episode logs.
  • Ingestion points: Historical transcript data from ~/.claude/projects/ and episodic memory notes from .claude/agent-memory/.
  • Boundary markers: The instructions do not specify the use of delimiters or warnings to the agent to ignore potentially malicious instructions embedded in the logs being analyzed.
  • Capability inventory: The skill can execute local Python scripts, read agent configuration files, and write Markdown reports to the project directory.
  • Sanitization: There is no evidence of filtering or escaping content extracted from transcripts before it is used to formulate improvement targets.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 4, 2026, 02:08 PM
Security Audit — agent-trust-hub — delivery-loop-retro