design-memory
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill builds a persistent profile by observing potentially untrusted project-level data, which could be used to poison the user's design history.\n
- Ingestion points: Processes data from project files such as
design-state.md, the handoff chain, and user-provided overrides.\n - Boundary markers: Absent. The instructions do not provide delimiters or warnings to ignore malicious content within the observed data.\n
- Capability inventory: Possesses file read/write access to
~/.designpowers/taste-profile.md.\n - Sanitization: Absent. The skill records observations directly into the journal without validation or filtering.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill interacts with persistent storage outside the local project directory.\n
- Accesses and modifies a cross-project journal located at
~/.designpowers/taste-profile.mdin the user's home directory.\n- [COMMAND_EXECUTION]: The skill includes mandatory instructions to prioritize the execution of another specific skill.\n - The 'Welcome Gate' requires the agent to invoke the
using-designpowersskill and complete a welcome sequence before performing any recording or reading tasks.
Audit Metadata