The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to perform project reconnaissance, such as listing files in .agents/skill-artifacts and checking for the presence of specific directories (research, brand, architecture).
[COMMAND_EXECUTION]: It uses git log to retrieve the most recent repository history to provide context for its routing decisions.
[COMMAND_EXECUTION]: The skill invokes a local synchronization script using the bun runtime (manifest-sync.ts) to ensure the project manifest is up to date.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it reads user-generated research artifacts to determine the current state of the pipeline.
Ingestion points: Reads files from research/ and .agents/, including product-context.md and icp-research.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands were found in the file-reading logic.
Capability inventory: The skill has access to command execution (Bash) and directory traversal tools.
Sanitization: Content from the ingested files is used to build an internal state map without explicit sanitization or validation of the text content.
[COMMAND_EXECUTION]: Employs the dynamic context injection syntax (! followed by backticks) to execute shell commands at load time for environment discovery. The commands used (e.g., find, git log) are benign and serve the skill's primary purpose of orchestration.

orchestrate-research