The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx @useclawlink/cli to download and run the ClawLink CLI tool directly from the NPM registry.
Evidence: npx @useclawlink/cli login in SKILL.md.
Context: The package @useclawlink/cli is a resource belonging to the vendor, ClawLink.
[COMMAND_EXECUTION]: The skill relies on shell command execution to perform all FreeAgent operations, including login, connection, and data manipulation.
Evidence: Multiple npx @useclawlink/cli commands in SKILL.md used for actions like create_invoice, list_accounting_transactions, and get_the_pl_summary.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from FreeAgent (e.g., transaction notes, contact details, invoice descriptions) which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Actions such as freeagent_list_accounting_transactions, freeagent_company_details, and freeagent_get_the_pl_summary fetch data from the external FreeAgent API.
Boundary markers: None explicitly mentioned in the skill instructions to delimit tool output from agent instructions.
Capability inventory: The skill has the ability to write data (create invoices, contacts, bank accounts) and execute shell commands via the CLI.
Sanitization: No specific sanitization or validation of the retrieved financial data is described.

freeagent