The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill directs the Archivist agent to dynamically generate a Python build script (build_capsule.py) and subsequently execute it using python3 to render PDF documents from session data.
[EXTERNAL_DOWNLOADS]: The skill performs runtime installation of several packages including weasyprint, pdfkit, and @mermaid-js/mermaid-cli from public registries (PyPI and NPM). It also includes a fallback mechanism that injects a Mermaid library script from a public CDN into the generated distribution web page.
[PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection due to its processing of untrusted external content.
Ingestion points: SKILL.md (Step 1: Capture Source Materials) and the Explorer agent utilize WebFetch and WebSearch to ingest data from user-provided URLs and files.
Boundary markers: Absent; there are no explicit delimiters or instructions to ignore embedded commands within the ingested content.
Capability inventory: The skill has broad capabilities including shell command execution (bash, pip, npx), dynamic script execution (python3), file system modification, and network access.
Sanitization: Absent; the content fetched from URLs is converted to markdown and processed directly by the agent team without validation or filtering.
[COMMAND_EXECUTION]: The skill uses bash commands to modify agent environment configurations, create session directory structures, and execute package management tools.

team-ideation