docs-page
Warn
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The example documentation provided in
example.htmlincludes a command for piped shell execution (curl -fsSL https://get.filebase.dev | sh). This pattern downloads and executes a script from an untrusted external domain directly into the user's terminal, promoting unsafe installation practices.\n- [CREDENTIALS_UNSAFE]: Theexample.htmlfile references a sensitive file path (~/.config/filebase/credentials) used for storing authentication tokens. Mentioning such paths in templates or documentation can lead to accidental exposure or targeting of sensitive data by malicious actors.\n- [INDIRECT_PROMPT_INJECTION]: The skill instructions inSKILL.mdrequire the agent to read and process external data sources, including aDESIGN.mdfile and a documentation brief. There are no instructions for sanitizing these inputs or using boundary markers, creating a surface for indirect prompt injection where malicious instructions in the input files could influence the agent's behavior or be reflected in the generated HTML output.\n - Ingestion points:
DESIGN.md, documentation brief (SKILL.md).\n - Boundary markers: None specified in the workflow instructions.\n
- Capability inventory: Generation of HTML artifacts (SKILL.md).\n
- Sanitization: None specified for the content processed from the brief or design files.
Audit Metadata