| name | investigative-report |
| description | Use when turning raw material — research, notes, interview answers, fragments, or investigation findings — into a sourced written report or article, or when shaping, drafting, restructuring, or tightening such a draft. Triggers: "write up the findings", "investigative report", "turn these notes into an article", "draft the report", "shape this into an article", "edit the article", "help me write this up".
|
| triggers | ["investigative report","write up the findings","turn these notes into an article","draft the report","shape this into an article","edit the article"] |
| category | writing |
| token_efficiency | high |
| dependencies | [] |
Investigative Report
Overview
Turn gathered material into a written case. Every claim is provenance-backed, and the
citation gate is tiered — hard for findings, soft for color. The report is the
investigation board, rendered linearly. One disciplined pass replaces the old
fragments → shape → beats → edit pipeline.
When to use
- Raw input (notes, interview answers, research, findings) must become a report or article.
- A draft needs shaping, structuring, or tightening.
- Claims must be sourced and attributable — not a casual prose dump.
Not for: throwaway writing with no sourcing needs, or a single quick paragraph.
The five phases
- Gather — mine the user first, then web-research to corroborate and extend. Every
fragment becomes a board node carrying
source / method / confidence / verify.
Never flatten user input and web research into one undifferentiated pile — each keeps
its origin.
- Shape — find the through-line; state it back; revise on the user's response. The
thesis may rest only on hard-tier (verified) findings.
- Structure — default: marshal the findings into an argument for the thesis (red
string = the logical spine). Optional beats mode: assemble as a narrative journey.
- Edit — tighten, restructure, cut.
- Cite — classify every claim by tier, enforce the gate, render provenance.
The citation gate
kind | Tier | Gate |
|---|
finding | hard | must be verified, else demote to color or cut |
hypothesis · question | soft | render with confidence flag |
color / vignette / inference | soft | render with confidence flag |
The thesis rests only on hard-tier findings. A would-be finding that isn't verified
is demoted to a hedged color statement or cut — never stated as fact.
Render convention
Every claim ends with ^[<source> · <confidence>]; verify is the expandable citation.
Hard reads clean (gate guarantees verified). Soft shows corroborated (amber) or
unverified (red). Standalone color goes in a > field note blockquote, confidence-tagged.
Provenance
Reads the board and the provenance record (BOARD.md, spiffe-provenance.md):
verify → citation, confidence → tier eligibility, source → attribution. Nothing
appears in the report that isn't already a node on the board.
Boundaries
Independent skill. It does not investigate (that's investigate) and does not interrogate
specs (interrogate-spec). It produces writing only. The QUESTION.md voice is
flavor, never a hard dependency — the skill works without it.
Red flags — STOP
- About to state an
unverified claim as fact → demote to color or cut.
- "This color is basically a finding" → if it isn't
verified, it's color.
- "The thesis is close enough on corroborated evidence" → the thesis needs verified findings.
- Merging user and web fragments "for flow" → each fragment keeps its
source.
All of these mean: respect the tier. The gate is the difference between a report and an
opinion piece.