research-synthesis
Execute the synthesis phase of an research project. Analyzes the hypothesis graph, determines report verbosity (simple vs. complex path), generates the final report, creates a prototype inventory, and finalizes the implementation plan in the brief. Use when concluding a research
Install
npx skills add https://github.com/v0lka/skills/tree/main/engineering/research-synthesis
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install v0lka-skills@llmmart
git clone https://github.com/v0lka/skills.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole v0lka/skills collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
Research Synthesis
Execute the synthesis phase: analyze results, generate the final report, consolidate prototypes, and finalize the implementation plan.
Context
Synthesis is the final phase of a research project. It begins when one of these conditions is met:
- Answer obtained — the research question received a definitive answer (positive, negative, or refined).
- Budget exhausted — allocated time or resources have run out.
- External decision — stakeholder or research director decides to conclude the research (e.g., after a quarterly review).
Procedure
Step 1 — Confirm synthesis readiness
Read the research brief (brief.md) and hypothesis graph (hypotheses/graph.md)
to determine which trigger condition applies. Present the researcher with a
summary:
- Trigger: why synthesis is starting (answer obtained / budget / external).
- Research question: original formulation and any refinements.
- Graph stats: total hypotheses, confirmed, refuted, cancelled, still open.
- Open fronts: any hypotheses still in
openorin-progressstatus.
If there are open hypotheses, ask the researcher whether to:
- Cancel them (they will be documented as "not pursued").
- Complete them before synthesis (defer synthesis).
Step 2 — Determine report mode
The methodology defines two report modes based on the complexity of the path taken:
Simple path — hypotheses mostly confirmed, few or no pivots/dead ends. Produces a minimal report (1–2 pages equivalent).
Complex path — significant pivots, dead ends, unexpected discoveries. Produces an extended report with graph analysis and decision documentation.
To determine the mode, evaluate:
- Ratio of refuted/cancelled hypotheses to total.
- Number of pivots (nodes where a branch was abandoned for a new direction).
- Whether any major forks occurred.
- Whether the final answer differs significantly from initial assumptions.
Heuristic: if more than ~30% of hypotheses were refuted or cancelled, or if there were 2+ pivots, use the complex report mode.
Present the assessment to the researcher and let them confirm or override the mode choice.
Step 3 — Analyze the hypothesis graph
Walk the hypothesis graph and identify:
- The critical path — the sequence of confirmed hypotheses that forms the backbone of the answer.
- Key decision points — where continue/pivot/kill/fork decisions significantly shaped the research direction.
- Notable dead ends — refuted or killed branches that produced valuable negative knowledge.
- Unexpected discoveries — findings outside the original scope that emerged from experiments.
Step 4 — Inventory prototypes
Scan all hypothesis cards for links to prototypes, PoCs, tools, scripts, and other artifacts. For each prototype:
- Name / description: what it is.
- Location: where the code/artifact lives (repository, branch, directory).
- Maturity level: minimal PoC / functional prototype / near-production.
- How to run: brief launch instructions (or reference to detailed docs).
- Which hypothesis it validates: link back to the card.
Determine which prototypes should be consolidated into a unified solution (if applicable) and which are standalone artifacts.
Step 5 — Finalize the implementation plan
Update the "Implementation plan" section of brief.md to contain:
- Target product: which existing product receives the results, or whether a new product is being created.
- Implementation model: one of:
- Development team integration — results are exhaustive, no pivots expected during integration. A working group (research expert + developers) implements in the standard development process.
- Research team integration — integration may involve uncertainty and pivots. A working group (development expert + researchers) proceeds using the research methodology.
- New product (MVP) — results form a new product. Researchers build the MVP; then hand off to development.
- Calendar plan: estimated timeline in development sprints or equivalent units.
- Effort estimate: order-of-magnitude effort in person-sprints.
- Key risks: what could go wrong during implementation.
The choice of model depends on the level of uncertainty: if pivots are expected during integration, use the research methodology; if the path is predictable, use the standard development process.
Step 6 — Generate the final report
Create report.md in the research project directory using the appropriate
template:
- Simple path: assets/report-simple-template.md
- Complex path: assets/report-complex-template.md
Simple report contains:
- Problem statement (1 paragraph).
- Approach description (1 paragraph).
- Key results.
- Prototype inventory (what it is, where it lives, how to run it).
- Implementation plan summary with chosen model.
Expected length: 1–2 pages equivalent.
Complex report additionally contains:
- Hypothesis graph with results (embedded Mermaid or link to graph.md).
- Description of key decision points and their rationale.
- Description of notable dead ends worth knowing about.
- Unexpected discoveries and their potential implications.
- Open questions that remain unanswered and may warrant future research.
Length: as much as needed to transfer all knowledge gained.
Guiding principle: write more only when there is something to explain. A successful straight path does not need justification.
Step 7 — Update the brief
- Set the Status field in
brief.mdto "Completed" (or "Completed — budget exhausted" / "Completed — external decision" as appropriate). - Set a completion date.
- Ensure the implementation plan is finalized (Step 5).
Step 8 — Update the research index
Update the entry in {research-root}/index.md to reflect the completed
status and add a link to the final report.
Step 9 — Record the conclusion to the research log
Append a best-effort note entry to log.md in the research project
directory (R-NNN/log.md) recording that synthesis completed (see
"Research Log" below). The entry is project-scoped: omit the hypothesis
identifier.
Quarterly Review Preparation
If synthesis coincides with a quarterly review/demo, prepare a presentation covering:
- Brief state (including changes since last review).
- Hypothesis graph with "front line" highlighted.
- Key findings and intermediate results.
- Prototype demonstrations (if applicable).
- Plans for the next period (or closure summary).
This is not a separate artifact — it is a summary derived from the report and graph.
Research Log
Maintain a research-history log at log.md in the research project directory
(R-NNN/log.md). When synthesis concludes the project, append one best-effort,
timestamped entry so the timeline fills in automatically.
Each entry is a level-2 heading plus a one-line message body:
## <kind> <created_at> <hypothesis-id>
<message>
<kind>—notefor the synthesis conclusion (this skill's entries).<created_at>— current UTC time in ISO 8601 (YYYY-MM-DDTHH:MM:SSZ, e.g.date -u +%Y-%m-%dT%H:%M:%SZ).<hypothesis-id>— omitted: the conclusion is project-scoped, not tied to a single hypothesis.<message>— a single line summarizing the conclusion, e.g.Synthesis complete: research question answered; report.md written.
If log.md does not exist, create it with a # Research Log heading first.
Always append — never rewrite existing entries. If the log cannot be written,
continue anyway; the log is best-effort and must not block the synthesis.
Relation to Other Skills
- Typically the last skill invoked in a research project lifecycle.
- Depends heavily on
research-hypothesis(graph data) andresearch-experiment(prototype inventory). research-statuscan provide the starting point for synthesis analysis.research-decisionmay trigger synthesis when it determines the research question is answered or budget is exhausted.
Files (skills)
-
assets
-
report-complex-template.md 1.7 KB
Loading… -
report-simple-template.md 718 B
Loading…
-
-
SKILL.md 8.5 KB
--- name: research-synthesis description: >- Execute the synthesis phase of an research project. Analyzes the hypothesis graph, determines report verbosity (simple vs. complex path), generates the final report, creates a prototype inventory, and finalizes the implementation plan in the brief. Use when concluding a research project, writing the final report, or when the research question has been answered or the budget is exhausted. metadata: domain: engineering methodology: iterative-engineering-research --- # Research Synthesis Execute the synthesis phase: analyze results, generate the final report, consolidate prototypes, and finalize the implementation plan. ## Context Synthesis is the final phase of a research project. It begins when one of these conditions is met: 1. **Answer obtained** — the research question received a definitive answer (positive, negative, or refined). 2. **Budget exhausted** — allocated time or resources have run out. 3. **External decision** — stakeholder or research director decides to conclude the research (e.g., after a quarterly review). ## Procedure ### Step 1 — Confirm synthesis readiness Read the research brief (`brief.md`) and hypothesis graph (`hypotheses/graph.md`) to determine which trigger condition applies. Present the researcher with a summary: - Trigger: why synthesis is starting (answer obtained / budget / external). - Research question: original formulation and any refinements. - Graph stats: total hypotheses, confirmed, refuted, cancelled, still open. - Open fronts: any hypotheses still in `open` or `in-progress` status. If there are open hypotheses, ask the researcher whether to: - Cancel them (they will be documented as "not pursued"). - Complete them before synthesis (defer synthesis). ### Step 2 — Determine report mode The methodology defines two report modes based on the complexity of the path taken: **Simple path** — hypotheses mostly confirmed, few or no pivots/dead ends. Produces a minimal report (1–2 pages equivalent). **Complex path** — significant pivots, dead ends, unexpected discoveries. Produces an extended report with graph analysis and decision documentation. To determine the mode, evaluate: - Ratio of refuted/cancelled hypotheses to total. - Number of pivots (nodes where a branch was abandoned for a new direction). - Whether any major forks occurred. - Whether the final answer differs significantly from initial assumptions. **Heuristic**: if more than ~30% of hypotheses were refuted or cancelled, or if there were 2+ pivots, use the complex report mode. Present the assessment to the researcher and let them confirm or override the mode choice. ### Step 3 — Analyze the hypothesis graph Walk the hypothesis graph and identify: 1. **The critical path** — the sequence of confirmed hypotheses that forms the backbone of the answer. 2. **Key decision points** — where continue/pivot/kill/fork decisions significantly shaped the research direction. 3. **Notable dead ends** — refuted or killed branches that produced valuable negative knowledge. 4. **Unexpected discoveries** — findings outside the original scope that emerged from experiments. ### Step 4 — Inventory prototypes Scan all hypothesis cards for links to prototypes, PoCs, tools, scripts, and other artifacts. For each prototype: - **Name / description**: what it is. - **Location**: where the code/artifact lives (repository, branch, directory). - **Maturity level**: minimal PoC / functional prototype / near-production. - **How to run**: brief launch instructions (or reference to detailed docs). - **Which hypothesis it validates**: link back to the card. Determine which prototypes should be consolidated into a unified solution (if applicable) and which are standalone artifacts. ### Step 5 — Finalize the implementation plan Update the "Implementation plan" section of `brief.md` to contain: 1. **Target product**: which existing product receives the results, or whether a new product is being created. 2. **Implementation model**: one of: - **Development team integration** — results are exhaustive, no pivots expected during integration. A working group (research expert + developers) implements in the standard development process. - **Research team integration** — integration may involve uncertainty and pivots. A working group (development expert + researchers) proceeds using the research methodology. - **New product (MVP)** — results form a new product. Researchers build the MVP; then hand off to development. 3. **Calendar plan**: estimated timeline in development sprints or equivalent units. 4. **Effort estimate**: order-of-magnitude effort in person-sprints. 5. **Key risks**: what could go wrong during implementation. The choice of model depends on the level of uncertainty: if pivots are expected during integration, use the research methodology; if the path is predictable, use the standard development process. ### Step 6 — Generate the final report Create `report.md` in the research project directory using the appropriate template: - Simple path: [assets/report-simple-template.md](assets/report-simple-template.md) - Complex path: [assets/report-complex-template.md](assets/report-complex-template.md) #### Simple report contains: 1. Problem statement (1 paragraph). 2. Approach description (1 paragraph). 3. Key results. 4. Prototype inventory (what it is, where it lives, how to run it). 5. Implementation plan summary with chosen model. Expected length: 1–2 pages equivalent. #### Complex report additionally contains: 6. Hypothesis graph with results (embedded Mermaid or link to graph.md). 7. Description of key decision points and their rationale. 8. Description of notable dead ends worth knowing about. 9. Unexpected discoveries and their potential implications. 10. Open questions that remain unanswered and may warrant future research. Length: as much as needed to transfer all knowledge gained. **Guiding principle**: write more only when there is something to explain. A successful straight path does not need justification. ### Step 7 — Update the brief 1. Set the Status field in `brief.md` to "Completed" (or "Completed — budget exhausted" / "Completed — external decision" as appropriate). 2. Set a completion date. 3. Ensure the implementation plan is finalized (Step 5). ### Step 8 — Update the research index Update the entry in `{research-root}/index.md` to reflect the completed status and add a link to the final report. ### Step 9 — Record the conclusion to the research log Append a best-effort `note` entry to `log.md` in the research project directory (`R-NNN/log.md`) recording that synthesis completed (see "Research Log" below). The entry is project-scoped: omit the hypothesis identifier. ## Quarterly Review Preparation If synthesis coincides with a quarterly review/demo, prepare a presentation covering: - Brief state (including changes since last review). - Hypothesis graph with "front line" highlighted. - Key findings and intermediate results. - Prototype demonstrations (if applicable). - Plans for the next period (or closure summary). This is not a separate artifact — it is a summary derived from the report and graph. ## Research Log Maintain a research-history log at `log.md` in the research project directory (`R-NNN/log.md`). When synthesis concludes the project, append one best-effort, timestamped entry so the timeline fills in automatically. Each entry is a level-2 heading plus a one-line message body: ``` ## <kind> <created_at> <hypothesis-id> <message> ``` - `<kind>` — `note` for the synthesis conclusion (this skill's entries). - `<created_at>` — current UTC time in ISO 8601 (`YYYY-MM-DDTHH:MM:SSZ`, e.g. `date -u +%Y-%m-%dT%H:%M:%SZ`). - `<hypothesis-id>` — omitted: the conclusion is project-scoped, not tied to a single hypothesis. - `<message>` — a single line summarizing the conclusion, e.g. `Synthesis complete: research question answered; report.md written.` If `log.md` does not exist, create it with a `# Research Log` heading first. Always append — never rewrite existing entries. If the log cannot be written, continue anyway; the log is best-effort and must not block the synthesis. ## Relation to Other Skills - Typically the last skill invoked in a research project lifecycle. - Depends heavily on `research-hypothesis` (graph data) and `research-experiment` (prototype inventory). - `research-status` can provide the starting point for synthesis analysis. - `research-decision` may trigger synthesis when it determines the research question is answered or budget is exhausted.
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.