The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes Salesforce debug logs which are external, untrusted data sources. Ingestion points: The skill fetches ApexLog bodies via REST API calls as described in SKILL.md. Boundary markers: No explicit prompt delimiters or ignore instructions are used when interpolating log content for analysis. Capability inventory: The agent has high-privilege capabilities including apex_execute (arbitrary code execution), tooling_api_dml (metadata/record modification), and sobject_dml (data modification). Sanitization: No sanitization of log content is specified before the AI analyzes it.
[COMMAND_EXECUTION]: The skill uses shell commands to detect environment capabilities and interact with the Salesforce CLI. Evidence: references/execution-modes.md uses commands such as command -v sf, python3 --version, test -f sfdx-project.json, and find ~/ -name to verify the execution mode and locate scripts.
[DYNAMIC_EXECUTION]: The skill utilizes tools that execute code at runtime on the platform or locally. Evidence: The apex_execute capability allows the agent to run arbitrary Apex code snippets in the connected Salesforce org for reproduction and testing. Evidence: references/execution-modes.md describes a mechanism where large tool responses include an artifactUrl that the agent is instructed to fetch and process locally.

sf-debug