The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from external community sources to determine graduation criteria and account standing.
Ingestion points: The skill processes pasted rule pages, moderator statements, DM screenshots, and analytics exports to build its plans.
Boundary markers: The instructions lack explicit delimiters (e.g., XML tags or special tokens) to isolate untrusted data from the instruction stream, though it does include a high-level security warning.
Capability inventory: The skill can execute local scripts (registry-events.py) to propose state changes and write plan documents to the local memory/ directory.
Sanitization: There is no technical sanitization or escaping of the ingested data; the skill relies entirely on the agent's ability to follow a meta-instruction to ignore potential overrides.
[COMMAND_EXECUTION]: The skill's operational logic involves calling local Python scripts located in the scripts/connectors/ directory (including discourse.py, hn.py, bluesky.py, and fediverse.py) to retrieve account and community data. It also invokes registry-events.py for proposing updates to the internal channel registry.
[EXTERNAL_DOWNLOADS]: The skill retrieves data from several well-known external community platforms and APIs, including Hacker News (via Algolia/Firebase APIs), Discourse forums, Bluesky, and the Fediverse. These operations are performed using keyless, public-read methods to gather account history and community rules.

participation-warmup-planner