The Agent Skills Directory

[SAFE]: The skill instructions define a legitimate and transparent developer workflow for feature implementation. The logic is focused on TDD and project verification using local tools.\n- [COMMAND_EXECUTION]: The agent is instructed to discover and execute existing project commands for testing and linting (e.g., npm test, pytest, cargo test). The skill also mentions using the bd (beads) CLI for local task tracking. These behaviors are consistent with the skill's stated purpose and leverage standard development toolchains.\n- [EXTERNAL_DOWNLOADS]: The skill references an optional external repository for experiment tracking (github.com/glebis/claude-skills). This is a documented vendor resource and does not involve automated or hidden downloads.\n- [SAFE]: Indirect Prompt Injection Surface: The skill ingests 'Goal Contracts' from local files and provides them to a 'fresh-context reviewer' agent for auditing. Because this content is typically authored by the local developer and used for process consistency checks, it represents a standard workflow rather than a high-risk vulnerability.\n
Ingestion points: docs/factory/*/goal.md (specified in SKILL.md)\n
Boundary markers: Absent\n
Capability inventory: Shell command execution (tests), file writing (logs)\n
Sanitization: Absent

feature-factory