session-retrospective-rules
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from external sources, specifically previous session transcripts and user message logs.
- Ingestion points: Reads transcripts from local folders such as
~/.claude/projects/or~/.codex/sessions/, and log files like.scratchpad/<slug>/loop-log.mdandprogress.md. - Boundary markers: The instructions do not specify any delimiters or safety markers when interpolating transcript content into the retrospective context.
- Capability inventory: The skill has the capability to write files (
.scratchpad/retrospective.md), execute shell commands (wc -l), and propose or apply project-wide rules inAGENTS.mdor new test scripts. - Sanitization: There is no evidence of sanitization or filtering applied to the ingested transcript data.
- [COMMAND_EXECUTION]: The skill uses shell commands and external tool invocations to perform its tasks.
- Evidence: It calculates skill size using
wc -l <skills folder>/*/SKILL.mdand suggests using platform-specific export commands likeopencode session export <id>to retrieve transcripts.
Audit Metadata