The Agent Skills Directory

[COMMAND_EXECUTION]: The skill discovers and executes shell commands for testing (TEST_COMMAND) and linting (LINT_COMMAND) by reading project configuration files such as package.json, Makefile, build.gradle, pom.xml, and Taskfile.yml.- [EXTERNAL_DOWNLOADS]: The skill accepts external URLs as input for requirements or design documents, which are passed to the ss-plan skill for processing.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the ingestion of external requirement documents and design wikis.
Ingestion points: External wiki pages, tickets, PRDs (via URLs), and plan files (docs/plans/).
Boundary markers: The orchestrator restricts implementer subagents to specific task text and designated file scope boundaries, avoiding providing the full plan file to sub-processes.
Capability inventory: The agent can execute git operations, shell-based test/lint commands, and dispatch subagents.
Sanitization: No explicit sanitization or filtering of external content is mentioned before it is interpolated into implementer prompts.

ss-coding