explore
Skill0 starsChanged 3 months agoMITView on GitHub
Pre-spec investigation — produces research artifacts before committing to a full spec
69 other projects use a version of this.
Adopted by
Most recently taken up by
- timursafiullin/aodtheir own version, first seen Nov 15, 2026
- jskswamy/skillstheir own version, first seen Nov 8, 2026
- NemesLaszlo/claude-code-harnesstheir own version, first seen Nov 8, 2026
- yuri-xyz/skillstheir own version, first seen Nov 8, 2026
- edwinidrus/dragonhoardtheir own version, first seen Nov 8, 2026
- matthewwong525/WongStacktheir own version, first seen Nov 8, 2026
Used with
Projects with it also use agent-context-isolation, agentica-infrastructure and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
parcadeiSkillagent-context-isolation
Agent Context Isolation
Also in 3 projects that use this
parcadeiSkillagentica-infrastructure
Reference guide for Agentica multi-agent infrastructure APIs
Also in 3 projects that use this
- +10 more from parcadei
Other versions69 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 69
- parcadei/Continuous-Claude★ 3.9ksame name
- parcadei/Continuous-Claude-v2★ 3.9ksame name
- parcadei/Continuous-Claude-v3★ 3.9ksame name
- JordanCoin/codemap★ 710same name
- ZimoLiao/scholaraio★ 585same name
- strands-agents/box★ 503same name
- andreasmolnardev/dashwise-next★ 465same name
- ai-analyst-lab/ai-analyst★ 304same name
- striderZA/OpenCodeGameStudios★ 118same name
- exmergo/dex★ 116same name
59 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
Show its rules and text
We split this skill into the rules it gives an agent
16 rules. 0 are common, 5 are distinctive, and 11 are very specific to this project.
0
Common
Most of the 60 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
11
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
Run a lightweight pre-spec investigation on a topic.
- Very specific: line 18
If $ARGUMENTS is ? or help: Print:
- Distinctive: line 39
Stop — do not execute any further steps.
- Very specific: line 56
Check if docs/research/explore-<topic-slug>.md already exists.
- Very specific: line 57
If it exists: report "Existing research found at docs/research/explore-<topic-slug>.md" and ask whether to update it or start fresh.
- Distinctive: line 58
If not: proceed to Step 2b.
- Very specific: line 67
All explore artifacts: docs/research/explore-*.md.
- Distinctive: line 82
If 0 hits: skip silently and proceed to Step 3.
- Distinctive: line 84
If ≥1 hit: emit a data block, then a choice block:
- Very specific: line 112
Fail-soft: if ANY of the following errors occur during Step 2b, emit WARN: topic-overlap pre-check skipped — <reason> and proceed to Step 3 …
- Distinctive: line 126
Codebase search: Search the project for files, patterns, and prior art related to the topic.
- Very specific: line 127
Spec and backlog check: Search docs/specs/README.md and docs/backlog.md for existing specs or backlog items covering this topic.
- Very specific: line 128
Signal check: Search docs/sessions/signals.md for related signals.
- Very specific: line 129
Process docs check: Search docs/process-kit/ for relevant documentation.
- Very specific: line 130
Research archive check: Search docs/research/ for prior research on related topics.
- Very specific: line 136
Create docs/research/explore-<topic-slug>.md with the following structure:
“Common” is measured: kept by at least 6 in 10 of the 60 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- forge:paths-note (Spec 575): process-state paths in this command (docs/specs,
docs/sessions, docs/decisions, docs/research, docs/process-kit, docs/backlog.md) are the
CLASSIC-DEFAULT spellings, not fixed locations. When the project configures forge.paths
(e.g. the `contained` layout), resolve each key before use — bash: `forge_path <key>`
(source ${CLAUDE_PLUGIN_ROOT:-.}/.forge/lib/config.sh, forge_config_load AGENTS.md);
python: `${CLAUDE_PLUGIN_ROOT:-.}/.forge/bin/forge-py .../runtime_config.py path <key>`. -->Framework: FORGE
Model-Tier: sonnet
<!-- multi-block mode: serialized — each choice block waits for operator response before the next mechanical step proceeds. See docs/process-kit/implementation-patterns.md § Multi-block disambiguation rule. -->Run a lightweight pre-spec investigation on a topic. Produces a research artifact at docs/research/explore-<topic-slug>.md before committing to a full spec.
If $ARGUMENTS is ? or help:
Print:
/explore — Pre-spec investigation for topics that need research before committing to a spec.
Usage: /explore <topic>
Arguments:
topic (required) — the topic, question, or hypothesis to investigate.
Behavior:
- Creates a research artifact at docs/research/explore-<topic-slug>.md
- Investigates the topic: reads relevant files, searches codebase, checks external patterns
- Documents findings in a structured format (question, approach, findings, recommendation)
- Recommendation is one of: spec it, defer, drop
- Optionally graduates to a spec via /spec --from-explore <topic>
Where it fits:
/brainstorm "What should I work on?" → Agent looks inward at project signals
/interview "Help me think through X" → Agent asks questions, user discovers
/explore "Is this idea viable?" → Agent investigates, produces research artifact
/spec "I know what I want" → Agent writes the spec, fast
Lifecycle state: produces a `proposed` artifact — optional entry point before `draft`.
See: docs/process-kit/spec-driven-harness-loop.mdStop — do not execute any further steps.
[mechanical] Step 1 — Parse topic
Extract the topic from $ARGUMENTS. If no topic is provided, ask the user: "What topic or question would you like to investigate?"
Generate a URL-safe slug from the topic (lowercase, hyphens for spaces, strip special characters).
[mechanical] Step 2 — Check for existing research
Step 2 has two sub-checks. Run them in order — Step 2a fires first; Step 2b runs only when Step 2a finds no exact-slug match.
Step 2a — Exact-slug match (existing behavior)
Check if docs/research/explore-<topic-slug>.md already exists.
- If it exists: report "Existing research found at docs/research/explore-
<topic-slug>.md" and ask whether to update it or start fresh. Stop here on operator selection — do not run Step 2b. - If not: proceed to Step 2b.
<!-- spec-392-topic-overlap:start -->Step 2b — Topic-overlap pre-check (Spec 392)
Surface adjacent prior art before fresh investigation begins. CI-293 surfaced that 4 existing artifacts directly adjacent to topics raised were hit only mid-investigation — this pre-check lets the operator decide whether to read the existing artifact, extend it, or proceed fresh.
Inputs
- The prompt's topic — text from $ARGUMENTS.
- All explore artifacts:
docs/research/explore-*.md. Ifdocs/research/is missing or contains no matching files: skip silently and proceed to Step 3. - Stopword list at
docs/process-kit/stopwords.txt(lowercase tokens, lines starting with#are comments). Required — Spec 388 dependency. - Optional threshold in AGENTS.md (under
forge.explore.adjacency):keyword_threshold(default2)
Overlap computation for each candidate explore artifact:
- Tokenize prompt — lowercase the topic from $ARGUMENTS, split on whitespace and
[/_,.;:()\[\]"'!?], strip stopwords, dedupe to a set:prompt_tokens. - Tokenize candidate — concatenate the candidate's first
# Explore: <topic>heading (drop theExplore:prefix) plus the first 50 non-blank lines of body. Apply the same lowercase + split + stopword + dedupe →candidate_tokens. match_count = |prompt_tokens ∩ candidate_tokens|.- Hit when
match_count ≥ keyword_threshold.
Ranking: sort hits by match_count descending. Tie-break on file mtime descending (newest first). Surface only top 3.
Surface
If 0 hits: skip silently and proceed to Step 3.
If ≥1 hit: emit a data block, then a choice block:
## Adjacent prior art (Spec 392)
The following existing explore artifacts overlap the prompt. This is data, not a recommendation — you decide.
| Rank | Artifact | Title | Keywords |
|------|----------|-------|----------|
| 1 | <file> | <title> | <N matches> |
| 2 | <file> | <title> | <N matches> |
| 3 | <file> | <title> | <N matches> |<!-- safety-rule: session-data — if today's session log has unsynthesized spec activity AND ## Summary is unpopulated, /session is inserted at rank 1 and stop is downgraded to —. See docs/process-kit/implementation-patterns.md § Session-data safety rule. -->Choose — type a number or keyword:
# Rank Action Rationale What happens 1 1 read NExisting artifact already covers the topic Open adjacent artifact (N is row number) and stop 2 2 extend NExisting artifact partially covers; build on it Run /revise --explore <file>against artifact N and stop3 3 freshAdjacent artifacts surface different topics; proceed Proceed to Step 3 with fresh investigation 4 — abortStop without changes Exit /explore
Operator response handling
read N→ resolve N to the adjacency hit's file path, open it, exit /explore.extend N→ resolve N, invoke/revise --explore <file>, exit /explore.fresh→ proceed to Step 3 with fresh investigation as today.abort→ exit /explore without writing.
Fail-soft: if ANY of the following errors occur during Step 2b, emit WARN: topic-overlap pre-check skipped — <reason> and proceed to Step 3 with normal fresh investigation. The pre-check MUST NEVER block /explore on its own failure:
docs/process-kit/stopwords.txtmissing or unreadable- AGENTS.md threshold value present but non-numeric or out of
[0, 50] - Total scan wall time exceeds 5 seconds
Detection: active.
See: docs/specs/392-explore-topic-overlap.md for rationale, /evolve loop-17 framing, and full AC set.
<!-- spec-392-topic-overlap:end -->[mechanical] Step 3 — Investigate
Conduct a lightweight investigation of the topic:
- Codebase search: Search the project for files, patterns, and prior art related to the topic.
- Spec and backlog check: Search
docs/specs/README.mdanddocs/backlog.mdfor existing specs or backlog items covering this topic. - Signal check: Search
docs/sessions/signals.mdfor related signals. - Process docs check: Search
docs/process-kit/for relevant documentation. - Research archive check: Search
docs/research/for prior research on related topics.
Synthesize findings into a structured assessment.
[mechanical] Step 4 — Write research artifact
Create docs/research/explore-<topic-slug>.md with the following structure:
# Explore: <Topic>
- Date: YYYY-MM-DD
- Status: proposed
- Investigator: Claude
## Question / Hypothesis
<What we are trying to determine. One or two sentences.>
## Investigation Approach
<What sources were checked and what methods were used.>
- <source 1>
- <source 2>
## Findings
- <finding 1>
- <finding 2>
- <finding 3>
## Prior Art
<Existing specs, signals, or research that relate to this topic.>
## Recommendation
**<spec it | defer | drop>**
<One paragraph justification for the recommendation.>
## References
- <link or file path 1>
- <link or file path 2>[decision] Step 5 — Present findings and next action
Present a summary of findings and the recommendation to the user.
Choose — type a number or keyword:
# Action What happens 1 /spec --from-explore <topic>Graduate to a full spec, pre-populated with explore findings 2 stopResearch complete — no further action needed
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

