# Dowsing Chain

> Emperor Time's ring-finger chain — intuition made procedure. Router for three aspects: task dowsing (find your own work from repo evidence), system dowsing (scan the machine for counterpart coding agents), and lie detection (flag contradictions between what is said and what is observed). Use when asked to "find your own tasks", scope vague work, or "see what agents are on this machine". Load one aspect file at a time per the Invocation Ritual.

- Skill: `arbi-elezi/dowsing-chain` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add arbi-elezi/dowsing-chain`
- Raw SKILL.md: https://api.skillmd.com/api/skills/arbi-elezi/dowsing-chain/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: arbi-elezi (https://skillmd.com/u/arbi-elezi)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/arbi-elezi/dowsing-chain

---


# Dowsing Chain — router

> *The pendulum doesn't guess. It swings toward what is actually there.*

Dowsing looks like intuition from the outside. Inside, it is a disciplined
sweep: every candidate task and every detected agent must be backed by observed
evidence. **You may not invent work.** A dowsed task with no evidence behind it
is a hallucination wearing a pendulum.

## Selection table — read exactly one

| Your situation | Aspect file |
|---|---|
| "Find something worth doing" / vague goal to scope / between tasks with standing self-assign authority / triaging a codebase | `task-dowsing.md` |
| Building the enlistable-agent roster / "what agents are on this machine?" / pre-Steal-Chain scan / an unknown CLI needs profiling | `system-dowsing.md` |
| A source contradicts another (docs vs code, comment vs behavior, agent claim vs observation, memory vs terminal) and you must rule | `lie-detection.md` |

Multiple situations at once → run them as separate selections in sequence
(task dowsing often *triggers* lie detection; finish the sweep, then rule on
the contradictions it surfaced).

## Chain-wide laws (apply in every aspect)

1. **Evidence per swing** — every finding cites a quote, a `file:line`, or a
   command output captured this session. "Seems like" findings are discarded
   before they reach any report.
2. **Read-only chain** — dowsing observes; it never fixes, installs, or signs
   in. Anything worth changing becomes a task candidate for the client (or the
   waterfall), never an in-place edit "while scanning".
3. **One artifact per session** — a dowsing run ends in exactly one output
   (candidate table / roster / ruling), recorded in the ledger. An unrecorded
   sweep is lifespan spent on nothing (Vow 6).
4. **Skipped probes are declared** — inapplicable or unrunnable probes are
   listed as skipped in the artifact, so absence of findings is never mistaken
   for absence of problems.