Lock It
In

--- name: lock-it-in description: Turns the workflow from this chat into a tested, reusable skill. Use when the user types LOCK IT IN or asks to save, capture or turn this workflow into a skill. --- # Lock It In When the user says LOCK IT IN, turn the workflow we just finished into a real, reusable skill. Run all 4 phases in order. Don't skip phases and don't summarize. Do the actual work. ## Phase 1: Read the whole conversation Re-read the entire conversation. Note: - What the user was actually trying to do (the core job) - Which prompts and approaches worked (kept, praised, built on) - Which were rejected (corrected, redirected, called out) - The final version of the output the user was happy with - The clarifying questions the user answered that made the output better - Edge cases or "watch out for X" moments Keep these notes to yourself for now. ## Phase 2: Ask 2 or 3 clarifying questions Ask the user 2 or 3 short questions they didn't realize they needed to answer, like: - "Should this always do X, or only when Y?" - "If [piece of context] is missing, should it ask or assume a default?" - "What's the one thing that should never happen when this runs?" Never ask more than 3, and never ask something the conversation already answered. The goal is to surface the rules the user has in their head but didn't say out loud. ## Phase 3: Write the skill Write the skill as a SKILL.md file using this structure: --- name: [lowercase-words-with-hyphens, max 64 characters] description: [What the skill does AND when to use it, including the words the user is likely to say. Under 200 characters.] --- # [Skill title] [1 to 3 sentences explaining the job] ## Inputs - [What the skill needs and where it comes from] ## Steps 1. [First action, as an instruction] 2. [Next action] ## Output format [Exactly what to hand back. Show the structure.] ## Examples ### Example 1 Input: [a realistic input] Output: [what the skill should produce] ### Example 2 (edge case) Input: [a tricky or incomplete input] Output: [the expected output] ## Rules - If [edge case], do [behavior] - Never [the thing from Phase 2 that should never happen] - Always [the thing that should always happen] ## Phase 4: Test it before handing it over Run the new skill on a made-up input that matches the workflow we just did. Check: - Does the output match the quality of what we built together? - Does it handle the edge cases we found? - Does it follow the never and always rules? If anything is off, fix the skill first. When the test passes, say "Locked it in." Then: 1. Show the full SKILL.md in a code block. 2. If you can create files, also save it as a ZIP containing a folder named after the skill with SKILL.md inside, ready to upload under Customize, then Skills. ## Rules - Never skip the Examples section. Include at least two, one of them an edge case. - Never skip Phase 4. An untested skill is just a sticky note. - If the user pushes back on the skill, treat it as new Phase 1 information and rewrite it. Don't argue.