pp-forkable

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (medium risk: 0.30). The skill's text describes automatic local journaling and unconditional background "teach" uploads, explicit instructions to import browser session cookies, and an outbound "webhook:" output sink — features that could weaken privacy or be misused to exfiltrate data, but it does not explicitly instruct malicious actions.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md, the runtime workflow ingests outsider-authored free text via the user’s natural-language question passed into forkable-pp-cli recall "<user's question>" --agent, which then drives subsequent discovery/teaching behavior.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill requires installing and running a remote Go module which fetches and installs executable code at runtime (e.g. "go install github.com/mvanhorn/printing-press-library/library/food-and-dining/forkable/cmd/forkable-pp-cli@latest"), so the github.com module reference is a runtime external dependency that executes remote code.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill exposes explicit write commands that place, confirm, skip, reorder, and otherwise mutate real Forkable meal orders — actions the docs state "place real orders and spend" (they are dry-run by default but apply when invoked with --confirm). The CLI requires user-auth session cookies and issues live GraphQL mutations for meal management, so it can cause on-account charges and change spending. These are domain-specific transaction actions (not generic tooling), so this is direct financial execution capability.