atlas-grounding
Maps feature seams and current system state to ground architectural planning and brainstorming sessions.
Install
mkdir -p .claude/skills/atlas-grounding && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/19181" && unzip -o skill.zip -d .claude/skills/atlas-grounding && rm skill.zipInstalls to .claude/skills/atlas-grounding
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Use when starting a planning or brainstorm pass on a feature or subsystem (or grounding external prior-art) and you need current ground-truth across the architecture's relevant seams before designing — "we're exploring X, ground my understanding", "map the seams for X", or as the grounding step before /brainstorm, /shift, or app-port.Key capabilities
- →Decompose a subject into atomic concerns
- →Locate relevant architecture seams using a concern-routing atlas
- →Fan out read-only grounding agents per seam
- →Synthesize a seam-map slice and grounded briefs
- →Verify load-bearing claims adversarially
- →Hand off grounded briefs to other skills
How it works
This skill decomposes a subject into atomic concerns, routes each concern through a seam catalog, and dispatches grounding agents to gather current reality from relevant seams. It then synthesizes the gathered information into a seam-map slice and grounded briefs.
Inputs & outputs
When to use atlas-grounding
- →Brainstorming new features
- →Planning cross-seam refactoring
- →Grounding prior art exploration
- →Initializing app-port tasks
About this skill
Atlas Grounding
Overview
Route a subject — an internal feature OR external prior-art — into the architecture's seam-space using the concern-routing atlas, fan out a read-only grounding agent per relevant seam, and synthesize a seam-map slice + grounded briefs. The atlas is the map; parallel grounding agents are the method. This is the agentic-architect search primitive that app-port composes for prior art.
When to Use
- Before a
/brainstormor/shifton anything spanning more than one seam. - "We're exploring X" — you need which seams X touches AND their current reality.
- As
app-port's grounding step over a prior-art repo + our substrate.
Not for: a single-file lookup (just read it); a concern that clearly lives in one known seam (go straight there).
The Atlas (the map)
genesis/docs/content/elohim-protocol/architecture/2026-06-21-elohim-seam-map-concern-routing.md — the seam catalog (§3), the concern-routing table (§4), the device spectrum (§2). The atlas is durable/positive; current build-state lives in the dated assessments in .claude/data/ (linked from atlas §6) — read those for "what's wired now."
Method
- Decompose, then locate. First split the subject into atomic concerns — one clause ≈ one concern. The atlas §4 routing table is keyed by single concerns, so a multi-clause subject ("syncs across devices AND survives offline AND stays private") must be broken down, or you'll match only the loudest concern and silently miss the rest. Route each concern through the §4 table + §3 seam catalog. List every seam touched (composition-stack + role seams + device rungs), and use each seam's "confusion-to-avoid" note (§3) while routing — to reject misroutes and bound the seam count (a durability concern wearing a hub costume; a dataplane fact read as a doorway one). Name your exclusions and why.
- Fan out one read-only grounding agent per seam. Each grounds current reality: what exists (file:line, from the seam's §3 Home) and its build-state — the dated assessment if the seam has one (atlas §6), else fall back to the seam's §3 Home source files and separate wired-vs-spec directly. §6 is NOT total — confirm each routed seam's build-state source before fanning out. Use prose returns, not schemas (schema'd agents hang on empty payloads).
- Verify the load-bearing claims adversarially — separate wired from spec-only — especially for build-state.
- Synthesize a seam-map slice (the subject's seams + how they compose, with the cross-seam couplings) + per-seam grounded briefs.
- Hand off to
/brainstorm,/shift, orapp-portwith the briefs as the grounded backdrop.
For ≥3 seams, run the fan-out as a Workflow (parallel ground → adversarial-verify pipeline → synthesis).
Quick Reference — the seams (atlas §3)
hardware · OS/packaging · runtime/footprint · mod/plugin · SDK grammar · bridge · app-manifest/domain · client · doorway projection · peer-hoster dataplane · aggregation · hub cluster ops · confidentiality · temporal · resource-governance. Tracks: T1 DHT-notary · T2 substrate · T3 spoke · T4 doorway.
Composition
app-portruns this on external prior-art (+ our seams) before decomposing.- Reads the dated assessments for current build-state (unlike
concept-mapping/app-port, which read the durable atlas). concept-mappingtranslates any conventional concept the grounding surfaces into its elohim seam + placement.
Common Mistakes
- Matching the whole subject to one §4 row — decompose into atomic concerns FIRST, or you'll catch sync and miss privacy + custody.
- Grounding from memory or source-reasoning instead of dispatching agents that read current reality.
- Routing to too many seams — use the §3 confusion-to-avoid notes to reject misroutes; name exclusions.
- Assuming atlas §6 lists every seam's build-state — it doesn't; for an unlisted seam, ground from its §3 Home source files.
- Schema'd grounding agents — they loop on empty payloads; use prose returns + a synthesis pass.
When not to use it
- →When performing a single-file lookup
- →When a concern clearly lives in one known seam
- →When matching a whole subject to one routing table row
Limitations
- →Requires decomposition of multi-clause subjects into atomic concerns
- →Grounding agents return prose, not schemas
- →The atlas's dated assessments are not total for build-state
How it compares
This skill automates the process of identifying architectural seams and gathering current reality from them, unlike manually searching through documentation or codebases.
Compared to similar skills
atlas-grounding side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| atlas-grounding (this skill) | 0 | 11d | No flags | Intermediate |
| openspec-explore | 4 | 5mo | Review | Beginner |
| thought-patterns | 6 | 7mo | No flags | Advanced |
| agent-planner | 3 | 5mo | No flags | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by ethosengine
View all by ethosengine →You might also like
openspec-explore
studyzy
进入探索模式 - 一个用于探索想法、调查问题和澄清需求的思考伙伴。当用户想要在进行更改之前或期间深入思考某事时使用。
thought-patterns
markpitt
Orchestrates comprehensive cognitive thinking patterns including sequential, systems, lateral, divergent/convergent, abductive, analogical, first-principles, double-loop learning, gestalt, metacognitive, and neurodivergent patterns (ADHD hyperfocus, autistic detail-orientation, dyslexic spatial reasoning). Analyzes tasks to select optimal pattern(s), chains multiple patterns when needed, and validates outputs before responding. Use for complex problem-solving, creative tasks, analytical challenges, or when diverse cognitive approaches enhance solution quality.
agent-planner
ruvnet
Agent skill for planner - invoke with $agent-planner
llm-council
am-will
Orchestrate a configurable, multi-member CLI planning council (Codex, Claude Code, Gemini, OpenCode, or custom) to produce independent implementation plans, anonymize and randomize them, then judge and merge into one final plan. Use when you need a robust, bias-resistant planning workflow, structured JSON outputs, retries, and failure handling across multiple CLI agents.
gathering-requirements
CaptainCrouton89
Systematically clarify user needs, preferences, and constraints before planning or implementation. Classifies work type, investigates existing systems, discovers edge cases and integration points, resolves assumptions, and creates detailed specifications. Use when building features, enhancements, or integrations where requirements need clarification.
project-development
muratcankoylan
This skill should be used when the user asks to "start an LLM project", "design batch pipeline", "evaluate task-model fit", "structure agent project", or mentions pipeline architecture, agent-assisted development, cost estimation, or choosing between LLM and traditional approaches.