strategy-consultant
Runs a full strategy consulting engagement end-to-end — from sharpening a vague business question into a decision-oriented problem statement
Open source Repository Open in the app JSON README (API)
About
Runs a full strategy consulting engagement end-to-end — from sharpening a vague business question into a decision-oriented problem statement, through parallel multi-agent research with independent validation, sense-checking for bias and analytical drift, storyline synthesis, to a sourced, executive-grade client report (.docx) ready for expert review. Designed to produce 80–90% of the insight as a high-quality working deliverable, so consultants can focus on the 10–20% that creates unique value: client judgment, proprietary insight, and sharpening the final recommendations.
Details
- Kind
- Plugins
- Topic
- Government & public data
- Publisher
- chipkorvyn
- Origin
- marketplace
- Category
- ferramentas
- Stars
- 4
- Forks
- 2
- Last push
- 2026-05-04T05:44:27Z
- Repository state
- ativo
- Language
- HTML
- License
- Apache-2.0
- Added
- 2026-08-30 01:48:58
- Updated
- 2026-08-30 01:48:58
- Origin id
chipkorvyn/strategy-consultant/strategy-consultant
README
# Strategy Consultant (a.k.a. Worker) A Cowork plugin that performs the analytical work of a strategy consultant or any knowledge worker. Takes a business problem from definition through research, validation, synthesis, and delivers an executive-grade deliverable. ## Entry point `/engagement` (backed by the `engagement-manager` skill) is the **singular entry point** to this plugin. It orchestrates problem-definition → research → sense-check → synthesis → storyline-critique → client-report → deliverable-validation as internal phases. The other three commands — `/research`, `/define-problem`, and `/interview-guide` — are **continuation-only**: they run when a namespaced engagement workspace (`./engagements/<slug>-<YYYY-MM-DD>/`) and its `engagement-state.json` already exist on disk. They also recognise legacy `./engagement/` workspaces from earlier versions for backward compatibility. If an agent pattern-matches a verb in a user's request to one of these sub-commands without an active workspace, the sub-command must stop and hand control back to `/engagement`, which creates the workspace and routes to the correct starting phase. Do not invoke sub-commands cold, and do not attempt to reconstruct upstream artifacts (Precision Anchor, Client Question Checklist, Source Material Extraction Log, data-source answers, engagement-state.json) inside a sub-command — that rebuilds the orchestrator piecemeal. ## Workspace and state Every engagement gets its own folder under `./engagements/`, e.g. `./engagements/europe-ev-charging-2026-05-03/`. Inside it, `engagement-state.json` is the source of truth for resume logic — it records the engagement id, current phase, completed phases, artifact paths, checkpoint approvals, the next required action, and timestamps. Every phase reads it before running and writes it after completing. Phase gates check `completed_phases` and required artifacts before allowing a phase to proceed; missing items route control back to engagement-manager rather than reconstructing context locally. See `skills/engagement-manager/SKILL.md` for the full schema and resume protocol. ## Who It's For - **Management Consultants** — Strategy and generalist consultants at firms who want to accelerate the research-to-deliverable cycle - **Corporate Strategy Teams** — In-house strategists running internal engagements for C-suite - **Corporate Development / M&A** — Teams scoping market entries, acquisitions, and competitive landscapes - **Independent Advisors / Boutique Firms** — Solo or small-team consultants who need the analytical horsepower of a larger team - **Any knowledge work** - Any work that requires similar skill ## How It Works This plugin is designed to handle 80–90% of the analytical workload in a strategy engagement — research, evidence validation, structured synthesis, and report drafting — so that the consultant can focus on the 10–20% that creates unique value: client relationship judgment, proprietary insight, and final recommendations. All outputs are intended as high-quality working drafts for expert review, not finished deliverables. ## What It Does This plugin runs a complete consulting-grade analytical workflow: 1. **Define the problem** — Sharpens a vague business challenge into a decision-oriented question with clear scope 2. **Structure hypotheses** (optional) — Builds a testable hypothesis tree when the problem benefits from structured decomposition 3. **Research** — Deploys two independent research agents in parallel, then validates findings with a third agent for consistency and source quality 4. **Sense-check** — Pressure-tests all findings through triangulation, steel-manning the counter-argument, math checks, and bias scanning 5. **Synthesize** — Builds a storyline that directly answers the client question, organized by the logic of the answer (not a rigid framework) 6. **Deliver** — Produces an executive-grade Word document with clear language, sourced claims, and zero consulting clichés ## Commands `/engagement` is the **only** fresh-start command in this plugin. The other three commands are **continuation-only** — they resume an active engagement and refuse to run cold. If a user asks for "research X" or "scope this problem" without an engagement workspace already on disk, the correct response is to invoke `/engagement`, not the matching sub-command. | Command | Type | Purpose | |---------|------|---------| | `/engagement <topic>` | **Fresh start (sole entry point)** | Run the full end-to-end consulting engagement | | `/research <topic>` | Continuation only | Resume the research phase of an active engagement | | `/define-problem <brief>` | Continuation only | Resume problem-definition inside an active engagement | | `/interview-guide <topic>` | Continuation only | Resume the expert-interview phase of an active engagement | **Why this matters.** When sub-commands appear as peers of `/engagement` on the command surface, agents pattern-match a verb in the user's request to a sub-command and skip the orchestrator — silently dropping problem definition, sense-check, synthesis, and the real delivery skill. Treating sub-commands as continuation-only preserves the phase sequence. ## Skills | Skill | Purpose | |-------|---------| | **engagement-manager** | Orchestrator — runs the full end-to-end workflow | | **problem-definition** | Scopes and sharpens the client question | | **hypothesis-tree** | Builds MECE hypothesis trees (optional, when it adds value) | | **research** | Dispatches two analysts in parallel, then research-validator validates their memos | | **sense-check** | Pressure-tests findings for robustness | | **synthesis** | Builds the client-facing storyline and argument | | **client-report** | Produces the final .docx deliverable | | **expert-interview** | Plans, guides, and processes expert interviews | ## Agents | Agent | Role | |-------|------| | **analyst-alpha** | Independent research thread A — direct angle (market data, financials, benchmarks) | | **analyst-bravo** | Independent research thread B — complementary angle (competitive dynamics, case studies, contrarian evidence) | | **analyst-deep** | Optional third research pass targeting under-explored sub-dimensions identified after the first two analysts | | **research-validator** | Phase 3 — cross-checks analyst memos for consistency, CS scoring, and Precision Anchor alignment (analyst-memo scope only) | | **synthesis-critic** | Phase 5.5 — independent critique of the storyline against precision-anchor, validated research, and sense-check | | **deliverable-validator** | Phase 6.5 — audits the final deliverable (.docx / .xlsx / .pptx) against all upstream artifacts and produces a Deliverable Gap Report | ## Connectors (Optional) This plugin works without any external integrations. See `CONNECTORS.md` for optional connectors (document storage) that enhance team workflows. ## Design Principles - **Delivery mode only** — produces outputs directly, no Socratic coaching - **Evidence-traced** — every claim in the final report traces back to a sourced finding - **Bias-resistant** — two independent research threads + validation reduces confirmation bias - **No forced frameworks** — the storyline structure follows the logic of the answer - **No clichés** — a comprehensive banned-language list ensures clear, direct executive writing - **Expert-in-the-loop** — designed as a 80–90% accelerator, not a replacement for professional judgment ## Entry-point doctrine for workflow plugins This section generalises the architecture used above so it is reusable when adding new skills to this plugin or when authoring other multi-phase workflow plugins. **The pattern** - **One orchestrator as the singular entry.** For a multi-phase workflow, exactly one skill (and its command) is the entry point. In this plugin: `engagement-manager` / `/engagement`. - **Sub-skills as continuation subroutines.** Every other command in the workflow runs only as a continuation of an existing workspace. It checks for the workspace on disk; if absent, it stops and routes back to the orchestrator. - **Gates owned by the skill that needs them.** Pre-flight gates (data-source inquiry, clarifying questions, etc.) live in the skill that needs the information, not in the entry-point command and not duplicated across siblings. They fire exactly once, in order, because the orchestrator invokes the skills in sequence. - **Namespaced workspace + a state file as the continuation test.** The orchestrator creates and maintains one folder per engagement under a well-known parent (here: `./engagements/<slug>-<date>/`), with a state file (`engagement-state.json`) that records phase progress, completed phases, artifact paths, and the next required action. Sub-commands check for the workspace and read its state file to decide whether to continue or hand control back to the orchestrator. Phase gates compare upstream artifacts and `completed_phases` against required inputs before running. **Why this matters.** When sub-commands appear as peers of the orchestrator on the command surface, agents pattern-match a verb in the user's request to a sub-command and skip the orchestrator — silently dropping problem definition, sense-check, synthesis, and the real delivery skill. Making the orchestrator the singular entry and demoting sub-commands to continuation-only preserves the phase sequence. **Things to avoid** - Do not bolt pre-flight gates onto each sub-command. That rebuilds the orchestrator piecemeal. - Do not remove sub-commands. They remain useful for continuations and for users who know exactly where they are in the workflow. - Do not weaken the enforcement language inside any skill's `SKILL.md` — the gates are authoritative. **Checklist when adding a new skill to this or any workflow plugin** 1. Does the skill belong inside an existing phase, or is it a new phase? If new, wire it into the orchestrator's phase list. 2. If it also needs a standalone command, mark that command `CONTINUATION COMMAND — do not use as a fresh entry point` at the very top (after frontmatter) and have it check for `./engagements/<slug>-<date>/engagement-state.json` before any other tool call (with `./engagement/precision-anchor.md` accepted as a backward-compatible fallback for legacy v1.4.0 engagements). 3. Put gates inside the skill's `SKILL.md`, not in the command file. 4. Add a leading sentence to the skill's `description:` frontmatter: "Internal phase of the strategy-consultant engagement — invoke via engagement-manager, not directly." 5. Update this README's command table and phase sequence.