The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves ingesting and processing untrusted data from external URLs and raw HTML, which creates an indirect prompt injection surface. Instructions maliciously embedded in a web article could potentially influence the agent's behavior during the clipping or transcription process.
Ingestion points: The skill reads article URLs and HTML content into the agent's context as part of the extraction workflow (SKILL.md).
Boundary markers: No explicit delimiters or boundary markers are instructed for the final note transcription to separate untrusted content from the agent's instructions.
Capability inventory: The skill utilizes create_note for writing to the filesystem, curl for network access, and makes authenticated POST requests to a vault API.
Sanitization: The skill recommends using "safe ASCII kebab-case" for filenames but lacks specific sanitization logic for the article body text against potential embedded prompt injection attacks.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using curl to fetch raw HTML and download article images to the local system.
[DYNAMIC_EXECUTION]: The skill provides Python script templates to be executed at runtime to parse HTML tags and associate images with captions using regular expressions. These scripts process data stored in temporary files.
[EXTERNAL_DOWNLOADS]: The skill downloads content, including HTML and images, from arbitrary external URLs as part of its intended archival process.
[DATA_EXFILTRATION]: The skill performs authenticated network write operations to a remote API endpoint (/api/v1/vaults/<vault_id>/attachments) to upload clipped images. While this is the skill's primary purpose, it involves sending data and credentials over the network to an external service.

clip-web-article