The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes a workflow where the agent processes and responds to external questions (e.g., from the Fondo CPA team), which could potentially contain malicious instructions intended to influence the agent's behavior.
Ingestion points: Questions received through the "authorized finance workflow" and "Fondo Dashboard" as described in the "Answering Fondo Questions" section.
Boundary markers: The instructions mention using authorized systems and access-controlled reports, but do not specify technical delimiters (like XML tags or specific prefixes) for untrusted user/third-party input.
Capability inventory: The skill is granted Read, Write, Edit, and Grep capabilities via the allowed-tools frontmatter.
Sanitization: The skill lacks explicit instructions for sanitizing or escaping the content of the questions it processes before acting upon them.
[SAFE]: The skill does not include any executable scripts, binaries, or code files; it is entirely composed of markdown instructions and configuration.
[SAFE]: The instructions contain a specific security control directing the agent to omit sensitive data, such as ledger rows, payroll, tax data, and credentials, from its output receipts.
[SAFE]: No network operations, remote code execution patterns, or persistence mechanisms were detected in the skill content.

fondo-core-workflow-a