Explore

Repository context. Gather first

Collect these with individual Bash calls, one command per call, never combined into a single invocation:

• Current branch, git branch --show-current
• Working tree status (empty = clean), git status --porcelain | head -20
• Project root, git rev-parse --show-toplevel

The pipe is the bound and belongs in the command. A read-time cap ("read only the first 20 entries") bounds nothing: the Bash tool returns the command's complete output into context before there is anything to decide about.

Treat a failure (not a repository, git unavailable) as an unknown value and carry on. Keep these as separate body Bash calls rather than pre-compute lines: the harness runs a skill's whole pre-compute block as one shell invocation, and a worktree-isolated session refuses a compound command that contains git. The dated record for that composition claim is the worktree skill's reference/gather-block.md, "The pre-compute block runs as one shell invocation".

These values orient this session only. The project root is an absolute machine path. Use it to resolve files while working, but never echo it into EXPLORE.md; the handoff artifact records relative paths (see the outcome gate below).

Routing. Dispatch by default

From the main conversation, this skill dispatches the discovery:explorer subagent. Exploration reads many files; keeping that out of the orchestrator's context window is the point. The agent loads the project's path-scoped rules, runs the six dimensions, writes the artifact set, and returns a bounded summary plus a file pointer, not the reads. The parent resolves the pre-dispatch envelope first, six fields (scope, reason, memory-slice path, memory root, budget, capability flags), written into the dispatch prompt as the labelled template in ${CLAUDE_PLUGIN_ROOT}/reference/parent-contract.md, not as prose the agent has to parse, and owns the post-dispatch boundary after: re-surfacing open_questions to the user, dispatching the sibling verifier, and writing its verdict back into EXPLORE.md, the explorer always returns verification: pending because it may not grade its own work, so an artifact left saying pending after the parent verified it cannot be told apart from one whose verifier never ran, and this artifact is the whole handoff a fresh session resumes from.

Run inline instead when any of these holds. Inline runs the identical workflow, and the escape hatch relaxes nothing:

• Tight turn-by-turn iteration. ≤~5 known files, findings feeding a same-session edit, and you will redirect as results land.
• Cost, a dispatched run pays the full six dimensions every time; a single-file question does not need an envelope.
• The invoking context is already a subagent. Dispatch-by-default is scoped to the main-conversation boundary. Hoisting, not nesting: the outer dispatch already supplied the fresh context, so a second hop only spends the inner agent's own window.

When selecting the dispatched route: probe check-dispatch-artifact.sh --help before dispatching; a denied or errored probe halts. An un-runnable post-dispatch gate is not a reason to take the inline escape hatch to dodge the gate. A legitimate inline run (tight iteration, cost, already-a-subagent) does not owe that script, it has no script verdict to self-grade, so do not apply this precondition to the inline path. Invocation forms: ${CLAUDE_PLUGIN_ROOT}/reference/parent-contract.md.

One named alternative: the built-in Explore subagent, for raw "where is X / how does Y work" search. Fast, read-only, context-isolated. It skips project memory (convention-blind) and neither runs this 6-dimension workflow nor writes EXPLORE.md. Pass key constraints in the prompt when conventions matter, and expect to write the artifact yourself. Scale 1→N by dispatching more, each owning a disjoint area.

Preload-liveness sentinel. A dispatched agent receives this body through its skills: preload, and a preload that fails to resolve is skipped silently. Logged to the debug log and nowhere else. The dated record for that harness behavior is ${CLAUDE_PLUGIN_ROOT}/reference/parent-contract.md, "Harness facts the dispatch design rests on". A dispatched run therefore echoes this token verbatim as preload_token in its return payload:

discovery-explore-preload-8e2b7d

A missing or mismatched token is a hard failure: the parent discards the run, never downgrades or accepts the artifact. Without it, a preload miss produces an undisciplined run that still writes an artifact. Indistinguishable from success at every other seam.

Post-dispatch acceptance gate. Parent-side, before the payload is believed. status: complete is the agent's claim about its own run, and a claim is not evidence. Grade the run off disk, against the memory-slice path from the parent's own pre-dispatch envelope, carry that path across the dispatch, because it is this gate's input, and never a path read out of the payload, because the failure this gate exists to catch is a payload that comes back carrying no pointer at all. In order:

Pre-dispatch: create the memory slice and touch <that slice>/.explore-dispatch as the gate's freshness baseline, then hand that file to the gate as --newer-than. Without it a slice that already holds an earlier run's artifact set passes every on-disk check even when this dispatch wrote nothing at all. Both shell forms of that one command are in ${CLAUDE_PLUGIN_ROOT}/reference/parent-contract.md ("The pre-dispatch baseline"). Copy the one matching this session's shell, because the POSIX form's touch is not a command in PowerShell and its directory flag is a parameter error there. Same file carries the envelope template this dispatch also owes. The memory root's self-ignoring .gitignore guard is a separate obligation this gate does not grade. Same file, "What this gate does not grade".

1. The payload is well-formed. preload_token matches the sentinel verbatim, and an artifact: pointer is present. Missing either is a failed dispatch whatever the status field says; a missing token is a discard, per the rule above.

   And scope_as_received matches the scope the parent actually sent. Compared against the envelope the parent wrote, not against what it meant. It is the only check here that fires on an input that is present and wrong. A mismatch is a failed dispatch: re-dispatch with the scope restated in a form that survives the trip (see the caveat under Scope); do not accept the artifact and mentally translate it. A well-formed payload carrying no scope_as_received is an out-of-date agent definition, not a pass.

2. The artifact set is actually on disk, and this run put it there:

   "${CLAUDE_PLUGIN_ROOT}/scripts/check-dispatch-artifact.sh" <the retained memory-slice path> \
     --index-name EXPLORE.md \
     --newer-than <that slice>/.explore-dispatch --expect-index <the payload's artifact: value>
   

   --index-name is required, not defaulted, the same gate grades /discovery:research runs, the two artifact families differ only in that name, and a gate that fails closed everywhere else must not guess which family it is looking at.

   The gate grades exactly the slice path it is given and never scans the slice for an index elsewhere, so on a collision the envelope's slice path is already the sub-slice the parent assigned pre-dispatch, and that same assigned path is the one handed here.

   Cite the exit status, 0 usable, 1 no usable artifact set, 2 ungradeable, not a reading of the directory, because the context most motivated to call the dispatch finished is the one that would be doing the reading. Only the slice path and --index-name are required, and that bare form is still a real gate: every optional check reports unchecked rather than passing quietly. Append --expect-sidecars <n> when the payload reported a sidecars: count, and drop any flag whose value the payload did not supply.

   The index= path in that output is authoritative downstream: the verifier's target and the handoff pointer both come from it, not from artifact:. pointer=mismatch means the payload named a file this gate never graded, a defect in the payload, not a naming preference to reconcile.

Any non-zero exit halts the workflow, and a gate that could not run at all is a FAIL, never a skip. An invocation above that is denied, prompts and is declined, or errors out halts exactly as a non-zero exit does; do not fall back to reading the directory. Do not proceed to research, planning, or an edit on the strength of an exploration that did not happen. Proceeding is the damage a silently-empty return actually causes; the missing artifact is only how it starts. Recovery ladder, and why a resume beats a re-dispatch: ${CLAUDE_PLUGIN_ROOT}/skills/explore/reference/dispatch.md.

One named exception, and it is an exception to the halt, not to the gate. Exit 1 with persistence: by-value in the payload means the agent finished and its environment refused every write, the one failure where a re-dispatch pays full price to reproduce the same refusal. There the parent writes the slice itself from the artifact bodies the payload carries verbatim, into the memory-slice path it resolved before dispatch (on that path the payload's artifact: value is a destination the agent names, never the anchor), and then re-runs the identical gate command above. The workflow proceeds only on a subsequent exit 0. If the second run is non-zero, the halt stands and the ladder resumes at the rung it was on. The freshness check needs nothing special: the parent writes after its own pre-dispatch touch, so the index is strictly newer than the baseline.

Read the by-value rung before performing that write: ${CLAUDE_PLUGIN_ROOT}/skills/explore/reference/dispatch.md. It carries the two conditions that bind the write (filename checking and the collision rule) and why a by-value payload of findings rather than artifact bodies is a failed dispatch rather than a fallback.

Coverage discipline when fanning out: (1) write a numbered gap-list before any deepen pass; (2) fan out by disjoint area, never split the six dimensions across agents; (3) whoever holds the workflow writes EXPLORE.md. discovery:explorer writes its own, while built-in Explore agents cannot write one at all, so their caller does.

Purpose

Read the code before changing it. This skill builds the local knowledge a change depends on: the code, its neighbours, its history, its tests, and the build and tool configuration that constrains the solution. The six dimensions below say what to read; the outcome gate says when it is enough.

Local counterpart to /discovery:research (external sources). Together: /discovery:explore for what IS, /discovery:research for what SHOULD BE.

Plan-mode for high-risk exploration (optional, inline only): when exploring unfamiliar code in a high-blast-radius area (security boundaries, critical infrastructure, code you might accidentally modify mid-investigation), switch into plan mode for harness-level read-only protection. Routine exploration of well-understood code does not need this. A dispatched run cannot switch into it. EnterPlanMode is filtered out of every non-fork subagent unconditionally, and ExitPlanMode is filtered from every non-fork subagent too, "unless the subagent's permissionMode is plan". The dated record for that harness behavior is ${CLAUDE_PLUGIN_ROOT}/reference/parent-contract.md, "Harness facts the dispatch design rests on". discovery:explorer lists neither tool in its tools allowlist, so it holds neither either way. There the read-only boundary is the agent's own instruction, honored deliberately rather than enforced by the harness.

Scope

Explore the following: $ARGUMENTS

A dispatched run does not read that line. The scope does not reach a preloaded body by argument substitution, and a non-fork subagent has no view of the conversation to fall back on, so do not rely on seeing an unfilled slot: for a dispatched run the scope arrives in the dispatch prompt, and its absence is a parent-envelope failure the agent reports rather than repairs, whatever the line above renders as. There is no unscoped orientation mode under dispatch: a general repository sweep would hand back a plausible artifact answering a question nobody asked. What is documented about that path, and what is not, in either direction, is recorded once in ${CLAUDE_PLUGIN_ROOT}/reference/parent-contract.md. Running inline with no scope supplied above, infer it from the current conversation context. Identify what area of the codebase is relevant to the task at hand and explore that.

Caveat, a ${CLAUDE_…}-shaped token in a scope may not arrive as you typed it, which is a different question from the paragraph above and not evidence for or against it. What was observed, what is documented, what is not, and the practical rule: ${CLAUDE_PLUGIN_ROOT}/reference/parent-contract.md ("A different question"). The scope_as_received echo-back in the acceptance gate is what catches it whichever way the substitution actually runs.

Exploration dimensions

Cover the relevant subset of these dimensions. Not all apply to every task. Use judgment about which matter for the current scope.

1. Codebase reading

Read the actual code before forming opinions about it.

• Targeted files. Read files directly relevant to the task. Use Glob by pattern, Grep by content
• Adjacent code. Read code that calls, is called by, or is structurally similar to the target. Understand the neighborhood, not just the target
• Existing patterns. Before proposing a new pattern, search for how the same concern is handled elsewhere in the repo. Reuse > reinvent (unless the existing pattern is an anti-pattern or well outside modern best practices and not documented as a pragmatic decision)
• Convention files. Check the consuming project's own CLAUDE.md and project rules for conventions that constrain the solution space
• Reference source as spec, when the task points at an existing implementation to match (a vendored library, another module, even another language), READ that source as the authoritative spec

2. Git history

Code has context only git reveals, who changed it, when, why, and what else changed alongside it.

• git log --oneline -20 <path>. Recent change frequency and commit style
• git log --oneline --all --since="2 weeks ago". Recent repo-wide activity
• git diff HEAD~5 -- <path>. What changed recently in the target area
• git blame <file>, when specific lines were last touched and by whom
• Missing files, when git status or history references files that don't exist on disk, they may be intentionally deleted, so do not open git archaeology on them unprompted. Inline, ask first. Dispatched, you cannot ask. Record the file as an open_questions entry for the parent to surface and move on; asking is not optional here, so proceeding anyway would silently violate the rule that protects a deliberate deletion

3. Project structure

Understand how the pieces fit together before moving any of them.

• Directory layout, if the project documents its repository structure, verify the doc matches reality; otherwise map the tree yourself
• Project references / imports. Map the dependency graph by grepping the ecosystem's import/reference token across its build-config files (per-ecosystem tokens: ${CLAUDE_PLUGIN_ROOT}/skills/explore/reference/ecosystem-discovery.md. Compose /toolchain:check's covered-ecosystem set when the toolchain plugin is installed, retaining fallback ecosystems the seam does not cover; otherwise the reference's fallback table)
• Solution / workspace membership. Check the repo's solution or workspace file at the root for what's included
• Layer boundaries. Respect any dependency-direction rules the project declares
• Planned direction. Cross-reference findings with any stated direction in the project's own CLAUDE.md or docs. Assess how changes must fit the repo's current state AND planned direction

4. Test discovery

Tests are executable documentation. They reveal intended behavior, edge cases, and existing coverage.

• Find test projects. Glob the per-ecosystem test patterns in ${CLAUDE_PLUGIN_ROOT}/skills/explore/reference/ecosystem-discovery.md (test-globs / test-content-grep are explore-owned even when composing the toolchain seam)
• Co-located tests. Check whether unit tests live next to their source (sibling test project, __tests__/, adjacent _test.go) or in a separate tree
• Cross-cutting tests, a repo-root tests/ for architecture, dependency, or naming-rule tests that span multiple libraries
• Test patterns. Read existing tests (start with 2-3, scale to the number of distinct patterns in play) to understand naming conventions, assertion style, and fixture patterns before writing new ones
• Coverage gaps. Identify areas with no test coverage that the current task touches

5. Configuration and build state

Build configuration constrains what's possible. Understand it before fighting it.

• Build configs. Read the ecosystem's build / package / config files per ${CLAUDE_PLUGIN_ROOT}/skills/explore/reference/ecosystem-discovery.md (explore-owned build-configs even when composing the toolchain seam; resolved project-discovery / anchor locate roots)
• Analyzer / lint config. .editorconfig for shared severity levels; ecosystem-specific analyzer/linter files
• Package versions. Check the ecosystem's manifest (lockfile + central-version-management file if applicable)
• CI/CD. .github/workflows/ (or the project's CI equivalent) for what's validated on every PR

6. Environment and machine state

When the task involves tooling, MCP servers, or infrastructure:

• Installed versions. Probe per ${CLAUDE_PLUGIN_ROOT}/skills/explore/reference/ecosystem-discovery.md (explore-owned runtime-version-cmd even when composing the toolchain seam; install-hint is install prose, not a version probe)
• MCP server status. Test with a read-only call before depending on it
• Worktree state. git worktree list, current branch, uncommitted changes
• Local config. Project-local settings for env vars and tokens (don't read secrets, just verify presence)

Exploration modes

The resolved scope shapes the exploration focus. Read from $ARGUMENTS inline, and from the dispatch prompt under dispatch, which is the only place a dispatched run gets it:

Multiple arguments combine: /discovery:explore payments deps tests explores that area's dependencies AND test coverage.

Surfacing the USER's unknown-unknowns before they work in unfamiliar territory, a better-prompt deliverable, not the EXPLORE.md artifact, is the sibling /discovery:blindspot skill.

Output format

Present exploration findings as:

1. Summary. 2-3 sentence overview of what was found
2. Current state. Key facts about the explored area (structure, patterns, dependencies). When the explored module has a domain-vocabulary or glossary file, frame findings using the module's domain vocabulary
3. Existing patterns. How similar concerns are handled elsewhere in the repo
4. Test coverage. What's tested, what's not, what test patterns are used
5. Constraints. Analyzers, conventions, layer rules, or CI gates that constrain the solution
6. Planned direction alignment. How findings relate to any direction the project documents
7. Open questions. Anything that needs clarification before proceeding, each with a one-line recommended default + escape hatch. Inline, surface these to the USER. Dispatched, return them as open_questions in the payload and the parent surfaces them. AskUserQuestion is filtered out of every non-fork subagent, so the payload is how they reach a human at all; the dated record is ${CLAUDE_PLUGIN_ROOT}/reference/parent-contract.md, "Harness facts the dispatch design rests on". Either way, silent downstream resolution of a surfaced open question is an anti-pattern; the hand-off changes, the rule does not

If invoked standalone, present findings directly. If invoked as part of a larger workflow, findings feed into subsequent research and planning steps.

Outcome gate (before EXPLORE.md handoff)

Before writing EXPLORE.md (or returning the summary), check the artifact against binary criteria read off it, not a "did I explore enough?" recap. Any FAIL → return to the named dimension and fix before handoff:

• Every Output-format section populated with specifics. Each of the 7 sections carries concrete findings, not placeholders or "TBD".
• Every load-bearing area covered OR listed as a numbered gap. Nothing the task plausibly depends on is silently unexplored.
• Conclusion-driving claims are Read-verified, not inferred from a filename or grep hit. Anything a downstream decision rests on came from reading the file or code.
• Paths are machine-agnostic. No finding in the artifact echoes an absolute machine path (notably the project root gathered above); every path it records is written relative to the repo root, or, when there is no repo root, to the current working directory, so the handoff stays portable across machines.
• Open questions handed off, never dropped. Surfaced to the user inline, or carried in the payload's open_questions for the parent to surface under dispatch. Each with a recommended default.

Final step: persist artifact for handoff

Write the exploration output to <memory_dir>/<slug>/EXPLORE.md, a memory-tier artifact, never committed. Destination, slug, and runtime guards resolve per the plugin's topic-docs binding (${CLAUDE_PLUGIN_ROOT}/reference/topic-docs.md).

This file is the authoritative stage summary, a fresh session must be able to resume external research or planning reading only this artifact.

EXPLORE.md is always an INDEX, at every size, not only past an overflow threshold. It carries a task restatement, a one-line abstract per sidecar copied verbatim from that sidecar's header, a section → file + anchor table, and the closing Next-stage-handoff naming what external research (/discovery:research) or planning needs. The 7-point Output format's content lives in sibling EXPLORE-<section>.md sidecars in the same directory, each opening with a machine-readable YAML header so a consumer can grep headers and read exactly one file. Schema and the two load-bearing placement rules. Sidecars stay inside <memory_dir>/<slug>/, and EXPLORE.md stays the entry point: ${CLAUDE_PLUGIN_ROOT}/skills/research/context/artifact-shape.md.

Sidecar bodies match their length to what the section needs. Cover the substance, but do not pad with filler sections, redundant summaries, or boilerplate; the index's one-line abstracts and the outcome gate's specifics are the floor, not an invitation to narrate.

Sidecar filenames are keyed on the SECTION, not the scope, a run has one scope and many sections, so a scope-keyed name gives every sidecar the same filename and later sections overwrite earlier ones. Use the same stable id the header's section field and the index anchor carry.

Sidecar headers use the EXPLORE schema, not the research one. Local evidence is a repository path and whether the file was actually Read. verified: read | grep | inferred, not a URL, a source tier, and a publishing pool. Handed the research header, a run either fabricates fields it has no values for or improvises a shape no consumer can parse; the fabrication is worse, because it launders a grep hit into the field a fetched primary would occupy. Schema and why verified is load-bearing: the artifact-shape spoke's "EXPLORE.md sidecar header" section.

If an unrelated EXPLORE.md already exists in that slice, do not clobber it, and do not rename the index to dodge it, since EXPLORE-*.md is the sidecar pattern and a renamed index collides with its own sidecars. Occupancy is the PARENT's to resolve, before any write: stat the slice root pre-dispatch, and when it is occupied assign a sub-slice <memory_dir>/<slug>/<scope-slug>/ as the envelope's slice path, so the whole artifact set is written there under its normal names. A worker never picks a sub-slice itself, and on an inline run this session is the parent and applies the same check before writing. A prior exploration lost to a filename collision is silent and unrecoverable.

Gotchas

• Fan-out without a numbered gap-list. Dispatching subagents before writing gaps produces duplicate reads and missed areas. The gap-list is the coverage-discipline gate.
• Handing off with placeholder sections. Every Output-format section needs specifics or an explicit numbered gap. "TBD" fails the outcome gate.
• Inferring from filenames without Read. Grep hits are discovery only; conclusion-driving claims need Read verification.
• Investigating deleted files without asking, when git status shows intentional deletes, ask before archaeology.

What this skill does NOT do

• Does not research externally. That's /discovery:research. This skill reads local code, git, and file system only
• Does not make changes. It explores. Execution is a separate step
• Does not make decisions. It presents what IS. The planning step decides what SHOULD BE
• Does not skip dimensions for "simple" tasks, a quick bug fix still benefits from reading the surrounding code and checking for tests
• Does not substitute for reading, when uncertain, Read the file. Don't infer from file names or git log alone