The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which creates a potential surface for indirect prompt injection.
Ingestion points: According to SKILL.md, the agent ingests data from external sources including niche descriptions, brand briefs, competitor creator lists, and exported candidate files to perform fit-scoring and draft outreach.
Boundary markers: The skill explicitly mandates that all agent-drafted outreach, briefs, and contracts must be reviewed and approved by a human operator before being executed through other channels.
Capability inventory: The skill can write data to the Busabase backend via busabase-provider.js and mark engagements as completed using the execute_decisions.mjs script. It relies on other skills (e.g., instagram-outreach, kelly-email) for actual message delivery.
Sanitization: The companion application uses an escapeHtml utility (defined in app.js and creator-views.js) to sanitize data rendered in the UI, and the skill instructions in SKILL.md emphasize storing only minimum necessary content for review.
[COMMAND_EXECUTION]: The skill involves the execution of local scripts and commands as part of its operational workflow.
Evidence: SKILL.md instructs the agent to run node scripts/execute_decisions.mjs --apply to finalize approved engagements.
Evidence: The README.md provides local development commands such as pnpm install and pnpm dev for the companion application.
[EXTERNAL_DOWNLOADS]: The skill requires downloading standard Node.js dependencies for its local operation and companion app.
Evidence: The package.json files in the root and content/kelly-creators-app/ list dependencies such as busabase-sdk, hono, and @hono/node-server. These are fetched from standard package registries during installation.

kelly-creators