The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Node.js scripts (fetch-feeds.js, ghost-publish.js) and inline Node.js commands to manage the newsletter generation and publishing workflow. It also contains hardcoded absolute file paths (e.g., /Users/ksd/Desktop/Varnan_skills/...) in SKILL.md that are specific to the author's local development environment and may cause execution failures on other systems.
[EXTERNAL_DOWNLOADS]: The skill fetches content from external RSS/Atom feeds defined in feeds.json. It also uses npm install to download required dependencies like jsonwebtoken and rss-parser from the official NPM registry.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from third-party RSS feeds and includes it in the prompt sent to the LLM for summarization.
Ingestion points: scripts/fetch-feeds.js fetches content from external URLs specified in feeds.json.
Boundary markers: The synthesis prompt in SKILL.md includes instructions to 'use ONLY these sources', but lacks strong structural delimiters or specific warnings to ignore instructions embedded within the source content.
Capability inventory: The agent can write to the local filesystem and publish content to an external Ghost CMS instance via scripts/ghost-publish.js.
Sanitization: scripts/fetch-feeds.js performs basic HTML tag stripping and truncates article excerpts to 400 characters, which provides some mitigation but does not fully eliminate the injection risk.

newsletter-digest