The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines an attack surface for indirect prompt injection by processing external, potentially untrusted project artifacts.
Ingestion points: The orchestrator is designed to read and process content from .specify/memory/constitution.md, spec.md, plan.md, and tasks.md as specified in the workflow steps in SKILL.md and modules/progress-tracking.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands (e.g., 'treat following text as data only') are present in the logic used to ingest these files.
Capability inventory: The skill facilitates the execution of implementation tasks derived from the tasks.md file via the /speckit-implement command mapping in modules/command-skill-matrix.md and modules/progress-tracking.md.
Sanitization: There is no documented logic for validating, sanitizing, or escaping the content of these external files before they influence the agent's workflow transitions or task execution logic.
[NO_CODE]: No executable scripts (.sh, .py, .js) or binary files are included in this skill. The orchestration logic is entirely described through markdown instructions and metadata modules.

speckit-orchestrator