Brief intake

brief-intake

You help the user (or the next skill in the chain) absorb raw unstructured input and emit a clean, structured brief shaped for whatever artifact comes next. This skill is the upstream collector — every other skilldrop generator opens with "Gather the essentials"; this is the skill that does the gathering once, so downstream skills can skip straight to drafting.

How to respond

1. Identify the downstream target. Ask if the user hasn't said: "What are you going to do with this brief?" Map the answer to one of:

   Downstream skill	Trigger phrases
   adr-generator	"ADR", "decision record", "capture a decision", "we decided X"
   design-doc	"design doc", "tech proposal", "one-pager for review", "RFC"
   runbook-generator	"runbook", "on-call doc", "operational doc", "prod-readiness"
   tech-comparison-matrix	"compare X vs Y", "vendor selection", "tech selection"
   exec-summary	"exec summary", "1-pager for the VP", "board update"
   deck-builder / slide-outliner	"deck", "slides", "presentation", "pitch", "review meeting"
   decision-log	"decisions and action items", "follow-ups from the meeting"
   (generic)	None of the above — emit the generic brief and let the user pick later

2. Ingest the source. The user will paste text, give a file path, or describe verbally. If the source is huge (>3000 words), tell them you're going to extract the load-bearing parts and skip the rest, then confirm scope before you start.

3. Extract once, into the target's shape. Each downstream skill needs a specific set of fields. Use the matching template:

   • templates/brief-adr.md — context · forcing function · options · chosen option · trade-offs
   • templates/brief-design-doc.md — what · who · why now · constraints · open questions
   • templates/brief-runbook.md — service · owners · stack · deployment surface · known failure modes
   • templates/brief-tech-comparison.md — question · candidates · evaluation criteria · constraints
   • templates/brief-exec-summary.md — ask · audience · business impact · what you need
   • templates/brief-deck.md — topic · audience · time budget · key messages · palette/brand
   • templates/brief-decision-log.md — source · scope · today's date
   • templates/brief-generic.md — fallback when target is unknown

4. Mark every field with a confidence tag. Each extracted field is tagged inline:

   • [explicit] — stated outright in the source
   • [implied] — clearly inferable from the source (quote the snippet)
   • [inferred] — your judgement, not in the source — the user should verify
   • [missing] — not in the source; the downstream skill will need to ask

   This is the audit trail. The downstream skill (or the user) can then trust [explicit] fields and challenge the [inferred] ones.

5. Surface the gaps explicitly. End with a "🚩 Missing for downstream" section listing fields that came back [missing] and which ones are load-bearing enough that the downstream skill will refuse to start without them. For example: "adr-generator will refuse to draft without at least 2 considered options — only 1 found in the source."

6. Don't draft the artifact yourself. This skill stops at the structured brief. Hand off explicitly: "Brief ready. Run /adr-generator with this brief as input."

Quality bar

• Quote, don't paraphrase, when extracting. When a field is [explicit] or [implied], include a 5–15 word verbatim snippet from the source so the downstream skill (and the user) can verify nothing was hallucinated.
• Convert relative dates to absolute dates. "Next Friday", "EOQ", "the sprint after this one" all become YYYY-MM-DD with (implied) if you had to compute it. Today's date is in the system context — use it.
• Don't invent people. If the source says "someone from platform should…", the owner is unowned, not a guess.
• Don't merge across sources unless the user asked. If the user pastes a Slack thread and a doc, ask whether you should treat them as one brief or extract separately.
• Strip noise aggressively. Greetings, off-topic banter, "lol", and emoji reactions get dropped. The brief is the signal, not the conversation.

When to use this skill

• ✅ Before invoking any other skilldrop generator when the input is messy.
• ✅ When a user pastes a 90-minute transcript and says "I need an ADR from this".
• ✅ When the next step's SKILL.md says "Gather the essentials" and the user has unstructured source material.
• ✅ As a checkpoint after decision-log — to turn a single extracted decision into an adr-generator brief.

When NOT to use this skill

• ❌ When the user has already given a clean, structured brief — go straight to the downstream skill.
• ❌ For pure summarization — that's not what this is. The output is shaped for a specific next step, not a generic summary.
• ❌ For raw transcript cleanup. If the user wants meeting notes cleaned up, use decision-log.

Anti-patterns to avoid

• ❌ Padding [inferred] fields to make the brief look complete. Empty [missing] is more honest and more useful.
• ❌ Restating the source verbatim. The brief is shaped, not a copy.
• ❌ Skipping the gaps section because the brief looks fine — the gaps are the most useful output for the user about to invoke the downstream skill.
• ❌ Asking 4+ clarifying questions. If the source is too thin, say so once and let the user decide whether to add material or proceed with gaps flagged.
• ❌ Emitting the brief in prose. Downstream skills want structured fields, not paragraphs.