The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches external content from user-provided URLs using curl and downloads the slop-detector package from the npm registry.
[REMOTE_CODE_EXECUTION]: The score.sh script utilizes npx to download and execute the slop-detector package (version 1.2.0) at runtime.
[COMMAND_EXECUTION]: Shell commands are executed to process input, including curl for network requests and sed for basic HTML stripping.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted text from external URLs and local files, creating a surface for indirect prompt injection attacks.
Ingestion points: Text fetched via curl in SKILL.md, files read through the Read tool, and user-pasted content.
Boundary markers: No delimiters or specific instructions are provided to the agent to treat the input text as data rather than instructions during the rewrite and analysis steps.
Capability inventory: The skill environment allows shell execution (Bash), network access (WebFetch/curl), and file system modifications (Write/Edit).
Sanitization: Employs a basic regex via sed to remove HTML tags from fetched content, but lacks robust sanitization to prevent malicious instructions embedded in the prose from influencing the agent's behavior.

slop-sense