planning-with-files
Optional persistent file-based planning for multi-phase or long-running work that benefits from recovery across sessions or…
Save agreed requirements and decisions as a durable implementation spec.
$ npx -y skills add khaneliman/khanelinix --skill to-spec --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/to-specContext preview
The summary Claude sees to decide when to auto-load this skill.
Save agreed requirements and decisions as a durable implementation spec.
name: to-spec description: Save agreed requirements and decisions as a durable implementation spec. license: Complete terms in LICENSE disable-model-invocation: true metadata: khanelinix-invocation-mode: user-only
Use only when the user explicitly requests this skill. Turn the current conversation or supplied source into a saved spec, not an implementation plan.
1. Read supplied documents and relevant repository guidance. Inspect enough code to distinguish current behavior from proposed behavior. 2. Preserve agreed outcomes, constraints, interfaces, rationale, non-goals, and verification decisions. Separate verified facts from assumptions. 3. Do not restart an interview or silently settle material product choices. Record unresolved choices and which outcomes they block. Ask only when the requested artifact cannot be meaningfully drafted without an answer. 4. Prefer existing test seams. Describe observable acceptance evidence and relevant prior tests without prescribing tests that merely mirror code.
Use [spec-template.md](references/spec-template.md) when writing the spec. Follow the requested path or an existing project convention. Otherwise use `.planning/<feature-slug>/spec.md` and report the chosen path. Keep each feature separate from unrelated active plans. Never write outputs into the skill package.
Give the spec a stable identity and revision. On an authorized update, preserve its identity, increment the revision, and identify changed decisions. Do not replace an unrelated existing file. Use concise requirements with stable IDs; scale their number to the task rather than padding a user-story inventory.
Include source references that survive a cleared conversation. Summarize any necessary conversation-only decisions in the artifact itself. Use repository revision and symbol/path hints for current-state evidence, not as permanent implementation mandates. Preserve exact interface or schema snippets when they express an agreed contract more precisely than prose.
Mark the spec `draft` while material choices remain unresolved. Mark it `ready` only when agreed scope and acceptance evidence suffice for decomposition. This status does not authorize implementation or external writes.
Read back the saved artifact. Check that every agreed requirement is represented, assumptions are labeled, references resolve, and verification matches the desired behavior. Report its path, revision, readiness, and any blocking choices.
Stop after the spec unless the user already requested a next step. Decomposition and implementation remain separate tasks. Do not invoke another user-only skill without the user's request.
Local output is the default. If the user requests GitHub publication, use `github-toolkit` when available; otherwise prepare the exact issue body and report that publication needs a suitable tool. Reuse existing publication authority, read back the published result, and retain its identifier in the local spec. Do not infer permission to commit, push, publish, or deploy from spec creation.
Run packages directly with: Or install from the packages output. For example:
Optional persistent file-based planning for multi-phase or long-running work that benefits from recovery across sessions or…
Audit coding-agent sessions and mine cross-harness history for recurring corrections, skill-routing gaps, and harness faults. Use…
Design and audit agent instructions, skills, plugins, subagents, prompts, hooks, and provider configuration. Use for AI-tool…
Sketch interfaces, state, and module structure before code. Use for explicit design-led implementation or as a caller-owned…
Spawn N parallel candidates at the same task, pick a base, graft the strongest parts of the losers into it. Use for /arena,…
Bevy workflows for ECS, migrations, first-party UI/input/rendering/assets/shaders, plugin selection, scenes, scheduling, hot…