memory-flush
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes local tools to manage state and retrieve data. It executes
python3 scripts/memory_prep.pyfor scan window computation and log rotation, and uses thegh(GitHub) CLI to fetch pull request information. These operations are integrated into the skill's logic for deterministic bookkeeping. - [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it processes data from potentially untrusted external sources.
- Ingestion points: The skill reads content from multiple log files in the
memory/logs/directory and retrieves PR titles and descriptions viagh pr list. - Boundary markers: There are no specific instructions or delimiters mentioned to isolate untrusted content from the agent's primary instructions during processing.
- Capability inventory: The agent possesses the capability to write to numerous files (
MEMORY.md,memory/topics/*.md,memory/logs/*.md) and execute shell commands (python3,gh). - Sanitization: The instructions do not specify any validation, escaping, or filtering of the content retrieved from logs or PRs before it is used to update the persistent memory index.
Audit Metadata