The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's initialization scripts download the Forge Static Review Tool (FSRT) binary from a vendor-controlled repository (atlassian-labs on GitHub). This is a standard procedure for this security auditing tool to ensure it uses the correct analysis engine.
[REMOTE_CODE_EXECUTION]: As part of its security review workflow, the skill downloads and executes a specialized static analysis binary. The execution is limited to the security scanner tool and is performed on the local environment to analyze the provided Forge app codebase.
[COMMAND_EXECUTION]: The skill requires the execution of shell scripts (run_static_analysis.sh or run_static_analysis.ps1) which manage the environment setup, download the necessary scanner components, and run the static analysis binary against the project root.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted user-provided source code from Atlassian Forge projects.
Ingestion points: The agent reads the contents of the target project directory, including manifest.yml and all source files in the static/ and src/ directories.
Boundary markers: The skill uses a structured two-tier rule loading model and explicit reporting mandates (source-to-sink tracing) to keep the analysis focused and prevent the agent from following instructions embedded in the analyzed code.
Capability inventory: The skill can execute shell scripts, run a local binary, and read the entire content of the provided project directory.
Sanitization: While the scripts do not perform explicit string sanitization, the workflow relies on the specialized fsrt binary for the initial scan and requires the agent to validate all findings with evidence before reporting.

forge-security-review