Menu
Produce a comprehensive, detail-preserving research report from all phase findings and agent evidence.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Conduct a structured discovery interview to produce a precise, confirmed ProblemBrief.
Delegate complex, multi-file coding tasks to an autonomous coding agent that can iterate and self-correct. Uses Claude Code if ANTHROPIC_API_KEY is set, otherwise falls back to Gemini via generate_code + execute.
Execute arbitrary Python code in an isolated subprocess. Use for data transformation, statistical analysis, geospatial processing, knowledge graph modeling, ontology engineering, RDF serialization, data visualization, document parsing, and any computation task. Always print() results to capture them.
Download a dataset from a URL and inspect its actual contents — schema, column types, shape, sample rows, and null counts. Supports CSV, Parquet, Excel, JSON, and JSONL. Use this to look inside a dataset, not just its landing page.
Download a PDF from a URL and extract its full text — ideal for research papers, reports, and documentation. Returns page count and extracted text.
Navigate to and extract content from specific web URLs. Good for reading data portals, documentation, and landing pages.
| name | report_synthesis |
| description | Produce a comprehensive, detail-preserving research report from all phase findings and agent evidence. |
| tools | [] |
| source | built-in |
| tags | ["synthesis","report","final"] |
Run once as the mandatory final step of every research run, after all phases have completed. Takes the full set of phase syntheses and every agent's evidence and produces a single long-form document — the primary deliverable of the run, written for a practitioner who needs to act on the research and will not go back to re-run it.
Do not summarise. Preserve and organise. Every piece of evidence, every data point, every finding from every agent belongs in this report. Your job is not to cut — it is to arrange all the detail into a structure that a reader can navigate. A reader should be able to make decisions based on this report alone without needing to dig into the raw evidence.
Structure the document as follows:
A specific title that reflects the actual research goal, not a generic label.
The single most important finding and its confidence level. What it means for the original goal. Anything a decision-maker must know upfront. This is the only brief section — write it last so it accurately reflects what follows.
For each phase, write a dedicated section with its full name as the heading. Within each phase section:
Patterns and conclusions that emerge across multiple phases. What phase 1 set up that phase 2 confirmed or contradicted. How the picture evolved. This is the only place where you synthesise across phases — the phase sections must preserve the raw findings in full.
Unknowns from the original brief that remain unresolved. Be direct and specific — do not paper over gaps or pretend the research answered everything.
What the research could not cover. Low-confidence areas and why. Data that was unavailable. Time-bound information that may have already changed.
Concrete, prioritised next steps grounded in specific findings from this report. Each recommendation cites the evidence it rests on. No generic advice — every recommendation must be traceable to something an agent found.
List every agent with their role, phase, confidence score, and their single most important contribution to the report.
A single markdown document. Starts with the level-1 title heading. Ends with Sources. No preamble. No code fences. No postscript. Every major factual claim has at least one [agent_id] citation where an agent produced it.
