The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an ingestion surface for external data, which is handled with appropriate safeguards.
Ingestion points: The skill reads data from Salesforce organization queries (sf data query) in SKILL.md and accepts user-provided verification codes for WhatsApp activation in references/phone-verification.md.
Boundary markers: The instructions use explicit logic and tables to parse specific JSON fields and validate the channel state before proceeding.
Capability inventory: The skill utilizes the sf CLI for Salesforce API interactions and performs local file writes to the /tmp directory.
Sanitization: User input for verification codes is sanitized by trimming whitespace and removing hyphens to ensure a 6-digit format before processing, as described in Step 3 of the phone verification reference.
[DATA_EXFILTRATION]: The skill uses static file paths in the /tmp directory (e.g., /tmp/amc-precheck-channel.json) for temporary storage of query results. The stored data pertains to Messaging Channel metadata and deployment status, which does not contain credentials or sensitive personal information. This is a common scripting pattern in development environments.
[SAFE]: The skill adheres to security best practices by utilizing the sf api request rest command. As noted in the 'Gotchas' reference, this manages authentication internally within the CLI transport, avoiding the unsafe practice of extracting OAuth access tokens into shell environment variables or temporary files.

service-de-channel-activate