Skill329 repo starsupdated 5d ago

formated-specs

Formated-specs generates a structured research design specification from a user-provided topic by mapping it through DARE's four-layer architecture (campaign, strategy, tactic, standard operating procedure). It outputs a machine-readable research_graph schema in JSON format documenting nodes representing each layer's components and edges showing their relationships. Use this skill when you need to translate a research objective into a clean, executable orchestration plan while ensuring the graph accurately reflects only the skills actually employed. The skill automatically concludes by loading formated-result to pair the specification with corresponding research outcomes in a single unified document.

View source Repository: de-anthropocentric-research-engine

Install in Claude Code

Copy

git clone --depth 1 https://github.com/yogsoth-ai/de-anthropocentric-research-engine /tmp/formated-specs && cp -r /tmp/formated-specs/self-iteration/2026-06-06-probe-pretrain/skills/formated-specs ~/.claude/skills/formated-specs

Then start a new Claude Code session; the skill loads automatically.

Definition

SKILL.md

# formated-specs

You are the DARE executor. The user (simulator) gives you a research topic. Following
DARE's 4-layer architecture (campaign->strategy->tactic->sop), produce the research
design spec and simultaneously emit a **clean research_graph**:

## Emit research_graph (machine-readable block written into the spec file)

Embed a fenced ```json graph block in the spec file, with structure:
- `nodes`: each = {id, layer in {campaign,strategy,tactic,sop}, skill_name, function}
- `edges`: each = {from, to, kind in {prereq, calls, produces}}

The graph must faithfully reflect the 4-layer orchestration you actually used; do not
fabricate skills you did not use.

## Hard constraints
- Do not edit any live DARE skill (you only *use* their capabilities to design).
- 4-layer invariant: do not add layers, do not merge layers.
- **Last step: you must `load formated-result`** -- load and run the formated-result
  skill to write research_result back into this spec file, so graph and result are
  same-source and paired.