| name | using-fauna-skills |
| description | Discovers and invokes Fauna skills. Use when starting a session or deciding which skill applies to the current task. This is the meta-skill that governs how all other skills are loaded via fauna_list_skills / fauna_get_skill. |
Using Fauna Skills
Overview
Fauna ships a small set of lifecycle skills that encode senior-engineering
discipline for autonomous tasks. Each skill is a workflow with explicit
verification — not a reference doc. This meta-skill tells you when to load
which one.
When to Use
- Starting a new autonomous task and unsure which skill applies
- A Kanban card is moving between columns and you need to know the gates
- The model is about to declare TASK_COMPLETE and you want to know what
evidence the verification gate expects
Discovery
Tasks arrive with a column or intent. Map to a skill:
Card in backlog → spec-driven-development (write acceptance criteria first)
Card in todo → planning-and-task-breakdown
Card in_progress → incremental-implementation + test-driven-development
Card in review → code-review-and-quality
Anything failing → debugging-and-error-recovery
The active skills for the current task are listed in your system prompt
under "ACTIVE SKILLS for this task". Load full bodies with
fauna_get_skill(name) — or fetch just one section with
fauna_get_skill(name, "Verification") to keep tokens low.
Process
These behaviors apply at all times, across all skills. They are
non-negotiable:
- Surface assumptions before acting on ambiguous requirements.
- Manage confusion — stop and name it rather than guessing through it.
- Push back when warranted — sycophancy is a failure mode.
- Enforce simplicity — if 100 lines suffice, do not write 1000.
- Scope discipline — touch only what you were asked to touch.
- Verify, do not assume — "seems right" is never sufficient.
Common Rationalizations
| Rationalization | Reality |
|---|
| "The skill is too generic for this task" | Skills are guardrails, not templates. Apply the discipline; skip steps that don't apply with a one-line note. |
| "I know how to do this without loading the skill" | Then loading it costs nothing. Load it anyway — the Verification section is what catches the mistakes you didn't think of. |
| "This task is too small for a verification gate" | Verification of a small task is cheap. Verification debt of a big task is expensive. |
Red Flags
- Declaring TASK_COMPLETE without citing evidence from the active skill's Verification checklist
- Loading the same skill repeatedly within one task (load once, refer back)
- Picking a skill that doesn't match the task's actual phase (e.g. shipping-and-launch for a debug task)
Verification
This is a meta-skill — it has no exit criteria of its own. Its presence is
verified by the autonomous loop: every TASK_COMPLETE goes through the
anti-rationalization gate, which reads the active skill's Verification
section and requires cited evidence for each item.
