The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied 'coin' and 'topic' strings which are interpolated into parameters for the news_feed_search_x and news_feed_get_social_sentiment tools.
Ingestion points: User-provided coin and topic parameters extracted during the Execution Workflow in SKILL.md.
Boundary markers: Safety rules in SKILL.md require neutral presentation, source attribution, and clear platform labeling, but no structural delimiters are used for the parameters.
Capability inventory: Parallel MCP tool calls for fetching qualitative discussion and quantitative sentiment data.
Sanitization: No explicit sanitization or escaping of the user-provided input parameters is documented.
[COMMAND_EXECUTION]: The skill includes a deterministic shell probe in SKILL.md (Step 0 — Wrapper routing probe) that executes when the skill is loaded to determine whether to delegate to a primary skill or fallback to the legacy workflow. It checks for the existence of files in several common IDE skill directories under the user's home path.
[EXTERNAL_DOWNLOADS]: The skill includes maintenance scripts (scripts/update-skill.sh and scripts/update-skill.ps1) that download updates from the vendor's repository at github.com/gate/gate-skills using git, ZIP, or tar.gz archives.
[DYNAMIC_EXECUTION]: The update mechanism in the maintenance scripts overwrites the skill's own instructions and scripts with remote content when the user confirms an update, allowing for the execution of new code in subsequent turns.

gate-news-communityscan