The Agent Skills Directory

[PERSISTENCE]: The skill utilizes a custom Python utility, scripts/profile_store.py, to maintain user context and event records across sessions. It persists data to a local JSON file at ~/.lovstudio/skills/profile.json or within the user's XDG_CONFIG_HOME. This is used for maintaining a history of event URLs and user-stated preferences. The script includes an explicit safety mechanism (SENSITIVE_PARTS) to block the storage of credentials, tokens, or other secret-like fields.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external content from huodongxing.com. This creates a surface for potential indirect prompt injection if malicious instructions were embedded in public event data.
Ingestion points: Data is read via ego-browser from Huodongxing event pages (Step 1 in SKILL.md).
Boundary markers: The skill targets the ativityJson JavaScript object, which provides structured data rather than raw, unstructured HTML prose, reducing the likelihood of accidental instruction obedience.
Capability inventory: The skill has browser read/write capabilities (ego-browser) and local file-write capabilities (profile_store.py).
Sanitization: The skill instructions emphasize reading specific schema-defined fields (e.g., Category, Tag) to perform diagnostics.
[DYNAMIC_EXECUTION]: The skill uses JavaScript injection in the browser context to hook the fetch API and read page variables (ativityJson). This is employed for diagnostic purposes to verify that form submissions include the necessary metadata to prevent platform-level data loss.
[EXTERNAL_DOWNLOADS]: The README.md provides installation instructions using git clone from the vendor's GitHub repository (github.com/lovstudio/publish-event-onto-hdx-skill). This is a standard setup procedure for a skill provided by a known developer and matches the author identity.

lov-publish-event-onto-hdx