The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several utility scripts (ingest_feedback.mjs, apply_clusters.mjs, execute_decisions.mjs) designed for data management and ingestion. These scripts operate on local JSON payloads or interact with Busabase via a dedicated SDK. They do not accept or execute arbitrary shell commands from user input.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-generated feedback from external sources like Discord, Slack, and email, which represents an attack surface for indirect injection.
Ingestion points: Data enters the system via scripts/ingest_feedback.mjs through JSON payloads constructed from external communication channels.
Boundary markers: Instructions require a multi-step triage and clustering process, followed by a mandatory human review phase in the companion UI before any roadmap changes are finalized.
Capability inventory: The skill facilitates writes to Busabase and can prepare hand-offs for other messaging skills, but it does not execute outbound actions directly.
Sanitization: The AirApp frontend utilizes a centralized escapeHtml utility to sanitize feedback text before rendering, effectively mitigating cross-site scripting (XSS) and injection risks in the dashboard.

kelly-feedback