compact-conversation

Pass

Audited by Gen Agent Trust Hub on Oct 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from the current conversation history and persists it to a local file, which creates a potential surface for malicious instructions to influence the agent in subsequent sessions that read the summary.
  • Ingestion points: Analyzes all messages in the current conversation history (SKILL.md).
  • Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions when writing the summary or reading it back.
  • Capability inventory: The skill requires the ability to write and update the file '.windsurf/conversation-summary.md' within the project directory.
  • Sanitization: No sanitization, escaping, or validation of the conversation content is performed before the data is written to the persistent summary file.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 31, 2026, 04:28 PM