session-retro

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements an automated feedback loop that is susceptible to Indirect Prompt Injection.
  • Ingestion points: The skill reads the entire conversation history to identify friction, mistakes, and knowledge (SKILL.md, Step 1).
  • Boundary markers: There are no markers or instructions defined to prevent the agent from following malicious instructions embedded in the conversation data it is processing.
  • Capability inventory: The skill has the authority to write to persistent configuration files like AGENTS.md, .claude/rules/, and CLAUDE.local.md, and to execute shell commands via the bd tool (SKILL.md, Step 2 and 3).
  • Sanitization: The skill explicitly instructs the agent to "Auto-apply all findings without asking for per-item approval," removing the human-in-the-loop and allowing injected content to become a permanent part of the environment (SKILL.md, Step 3).
  • [COMMAND_EXECUTION]: The skill uses extracted content from the conversation to run shell commands (bd remember and bd create). If the extracted "finding" contains shell metacharacters or malicious payloads, it could lead to unauthorized command execution within the beads environment (SKILL.md, Step 3).
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 03:50 PM