The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources and local files into temp/source.html. There are no explicit boundary markers or instructions to ignore embedded prompts within the source HTML during the conversion process. The agent's capabilities (Bash, Write) could be exploited if malicious instructions are hidden in the source markup.
Ingestion points: temp/source.html (loaded from URLs or local paths).
Boundary markers: Absent; the agent is not instructed to disregard instructions inside the HTML.
Capability inventory: Bash, Write, and Read tools; execution of local scripts and npm commands.
Sanitization: Absent; the raw HTML is parsed for component generation.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a local utility script (scripts/fetch-stitch.sh) for downloading content, and standard development commands such as npm install and npm run dev to verify the environment. These commands are consistent with the skill's purpose but represent the capability surface for an injection attack.
[EXTERNAL_DOWNLOADS]: The skill fetches HTML content from user-provided or API-provided URLs using curl via the fetch-stitch.sh script. While the script includes hardening such as retries and User-Agent spoofing to ensure reliability, it facilitates the retrieval of external data into the local workspace.

stitch-svelte-components