brief
/brief
Zamienia jedno pytanie decyzyjne w brief decyzyjny z dowodami, w którym każde twierdzenie ma źródło, kontrdowody są podane, a brakujące dowody nazwane, na lokalnym folderze bez konfiguracji MCP.
- Autor
- mshadmanrahman ↗
- Repozytorium
- pm-pilot ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 02.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 02.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Zamienia jedno pytanie decyzyjne w brief decyzyjny z dowodami, w którym każde twierdzenie ma źródło, kontrdowody są podane, a brakujące dowody nazwane, na lokalnym folderze bez konfiguracji MCP.
Wyzwalacze z opisu
Oryginalny opis (EN)
The front door to PM Pilot. Turns one decision question into an evidence-backed decision brief where every claim carries its source, counter-evidence is stated, and missing evidence is named rather than papered over. Works on a local folder with no MCP setup. Triggers on: "brief", "/brief", "should we", "decision brief", "help me decide", "make the case for", "what does the evidence say about".
Jak działa
- When to Activate
- The contract
- Execution
- Output format
- Rules
- Integration with other skills
Podgląd SKILL.md
--- name: brief description: | The front door to PM Pilot. Turns one decision question into an evidence-backed decision brief where every claim carries its source, counter-evidence is stated, and missing evidence is named rather than papered over. Works on a local folder with no MCP setup. Triggers on: "brief", "/brief", "should we", "decision brief", "help me decide", "make the case for", "what does the evidence say about". origin: pm-pilot version: 1.0.0 --- # Brief One command, one decision, one reviewable artifact. ``` /pm-core:brief "Should we prioritize bulk import for mid-market?" ``` This is the first thing a new user should run. It needs no MCP servers, no configuration, and no knowledge of the other skills. It reads a folder of evidence, produces a decision brief, and forces a decision state at the end. ## When to Activate - User runs `/pm-core:brief` with or without a question - User asks "should we X", "help me decide on X", "make the case for X" - User asks what the evidence says about a decision they are facing - A new user has just installed PM Pilot and does not know where to start ## The contract A brief is not a summary. It is an argument with its sources attached, and it is only useful if a reader can check it. So: - Every consequential claim names where it came from. - Every direct quote is verified against its source before delivery. - Evidence that cuts against the recommendation is stated, not omitted. - Evidence that does not exist is named as missing, not inferred. - The brief ends in a decision state, not in a draft. If you cannot meet all five, say which one failed and why. Never deliver a brief that looks complete when it is not. ## Execution ### Step 1: Name the decision Take the decision from the user's argument. If they ran the command bare, ask once: > What decision are you trying to make? One sentence, phrased as a question. Reject a topic disguised as a decision. "Bulk import" is a topic. "Should we prioritize bulk import for mid-market this quarter?" is a decision. If the user gives a topic, offer the sharpest decision question you can infer from it and ask them to confirm or correct it in one line. Do not proceed past a vague decision. Everything downstream inherits the framing. ### Step 2: Confirm sources, once Look for evidence in this order and show the user what you found: 1. The current folder and any `evidence/`, `research/`, `interviews/`, `notes/`, `transcripts/` subfolder 2. `context/product.md`, `context/company.md`, `context/personas.md`, `context/competitors.md` if they have content 3. Connected MCP tools, if any are available (Jira, Slack, Confluence, Granola, GitHub) Report it as a short list and ask one question: > Found 14 files in `research/` and 3 context files. Jira and Granola are connected. Use all of these, or point me somewhere else? Accept a yes and move. Accept a path and use it. This is the only confirmation gate in the whole flow. Do not ask again later. If nothing is found, say so plainly and ask for a folder or a paste. Never invent a source, and never produce a brief from the model's own background knowledge while implying it came from the user's evidence. ### Step 3: Gather and verify Route to the skills that already exist rather than duplicating them: | Evidence type | Skill to use | |---|---| | Customer interviews and transcripts | `/pm-discovery:interview-snapshot`, then `/pm-core:synthesize-interviews` | | Cross-tool history on a topic | `/pm-core:deep-context` | | Competitive and market claims | `/pm-content:market-research` | | A messy causal question underneath the decision | `/pm-core:issue-tree` | Apply the hallucination guard from `synthesize-interviews` to every direct quote regardless of which path produced it. Search the source for at least 60% of the quote's words, exact match, case insensitive. Flag any miss inline with `[UNVERIFIED - edit before citing]` and keep going. Never silently repair a quote. ### Step 4: Write the brief Use the output format below. Keep it to one screen for the body. Depth belongs in the evidence table, not in the prose. ### Step 5: Force a decision state A brief that ends as a draft has not done its job. Close by asking the user to pick one: - **Decide.** The evidence is sufficient. Record the call and the reasoning. - **Defer.** Right call, wrong time. Record what has to change for this to come back. - **Experiment.** The uncertainty is testable. Record the smallest test that would settle it. - **Research.** The evidence gap is real. Record exactly what is missing and who can get it. Write their answer into the brief before saving. A brief with no decision state is incomplete. ## Output format ```markdown # Decision Brief: <the decision question> **Date:** YYYY-MM-DD **Sources reviewed:** <n files, n tools> **Quotes verified:** 12/13 verified against source ## Recommendation <Two sentences. The call, and the single strongest reason for it.> ## The case for - <Claim.> [source: interviews/acme-2026-08-14.md, verified] - <Claim.> [source: Jira INS-3128, verified] ## The case against - <Claim that cuts against the recommendation.> [source: ...] State at least one, and mean it. If the evidence genuinely runs one way, write "No counter-evidence found in the reviewed sources" and treat that as a warning sign about the sources rather than a strength of the argument. ## What we do not know - <Missing evidence, and what it would take to get it.> - <Question no source in this set can answer.> ## Confidence <One of: high, medium, low.> <One sentence naming what would move it.> ## Decision **State:** decide | defer | experiment | research **Call:** <what was decided> **Revisit when:** <trigger> ``` Save to `briefs/YYYY-MM-DD-<slug>.md` unless the user says otherwise. ## Rules - One confirmation gate, in step 2. Everything else runs without asking. - Never present an inference as a sourced claim. An inference is labelled as yours. - A claim you cannot trace to a source in the confirmed set does not go in the brief. It goes in "What we do not know". - Do not soften a weak evidence base with hedged prose. Say the base is weak and name what is missing. - Web search is allowed for market and competitor claims, and those get their URL in the source tag like any other claim. - If the user's own prior decisions contradict the recommendation, surface that in "The case against". ## Integration with other skills - **Upstream**: nothing. This is the entry point. - **Downstream**: `/pm-core:prd` turns a decided brief into a spec. `/pm-core:critique` pressure-tests the brief before it goes to a stakeholder. `/crucible:crucible` for a decision big enough to deserve a council. - **Sibling**: `/pm-core:prioritize` when the question is a ranking across many items rather than one call.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · mshadmanrahman/pm-pilot | 100 | MIT |
| sekretmaksym-dev/design-skills | 100 | MIT |
| digital-stoic-org/agent-skills | 100 | MIT |
| wrg32786/aigent-os | 100 | MIT |
| magnifito/domain-name-engine | 100 | MIT |
| alirezarezvani/claude-skills | 100 | MIT |
| anthropics/knowledge-work-plugins | 100 | Apache-2.0 |
| educlopez/ui-craft | 100 | MIT |
| oliver-kriska/claude-elixir-phoenix | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki a698ee559c93 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 90research-brief
Bada pytanie i pisze krótki raport ze źródłami, gdy trzeba ustalić stan dowodów przed decyzją. - 100research-brief
Zbiera dowody przed decyzją, podsumowuje fakty z repozytorium lub źródeł zewnętrznych i przygotowuje brief przed specyfikacjami, planami i implementacją. - 100decision-question
Prosi użytkownika o wybór tylko wtedy, gdy praca zależy od jego decyzji, i przedstawia najmniejszy kompletny brief decyzyjny. - 90decision-brief
Przygotowuje notatkę decyzyjną: pytanie, sprawdzone fakty, założenia, opcje z kompromisami i rekomendację dla decydenta kierownictwa. - 100decision-question
Pyta użytkownika o wybór tylko wtedy, gdy blokuje go decyzja należąca do niego, i podaje najmniejszy kompletny opis wyboru. - 95create-brief
Zadaje pięć pytań po kolei i zamienia niejasny pomysł na produkt w brief produktowy, bez specyfikacji technicznej.