blue-ocean-strategy
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill includes an 'Emit Outcome Sidecar' section in
SKILL.mdthat instructs the agent to write session metadata and performance metrics (timestamp, action counts, and status) to a persistent file at~/.claude/skill-analytics/last-outcome-blue-ocean-strategy.json. While the logged data is telemetry-focused rather than sensitive user content, this persistent local file-writing behavior is unprompted and targets a hidden system directory. - [INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface by processing untrusted industry or company names provided by the user. (1) Ingestion points: The
[industry/company]parameter in the/blue-ocean analyzecommand inSKILL.md. (2) Boundary markers: Absent; there are no instructions to use delimiters or ignore potentially malicious instructions embedded in the input. (3) Capability inventory: File-system writing (via telemetry sidecar) and standard prompt generation. (4) Sanitization: Absent; the skill lacks validation or filtering logic for external inputs before they are interpolated into analysis prompts.
Audit Metadata