Exécutez n'importe quel Skill dans Manus
en un clic

Exécutez n'importe quel Skill dans Manus en un clic

explore

Étoiles78

Forks5

Mis à jour10 juin 2026 à 13:58

DEFAULT for multi-file codebase exploration ("how does X work / where is Y", tracing an unfamiliar subsystem, log analysis). Delegate the read-only sweep instead of self-grepping — protects your context. Never delegate design/synthesis. Self-check ONLY the most imporant load-bearing files/lines; delegate breadth you wouldn't otherwise open.

Installation

Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.

Exécuter dans Manus

Source

DawidWraga

DawidWraga/davstack

Ouvrir le dépôt GitHub Voir les dépôts du créateur

Téléchargement

Exécuter dans Manus

Métiers associésSOC

Basé sur la classification professionnelle SOC

Développeurs de logicielsProfessions informatiques et mathématiques·SOC 15-1252

SKILL.md

readonly

name	explore
description	DEFAULT for multi-file codebase exploration ("how does X work / where is Y", tracing an unfamiliar subsystem, log analysis). Delegate the read-only sweep instead of self-grepping — protects your context. Never delegate design/synthesis. Self-check ONLY the most imporant load-bearing files/lines; delegate breadth you wouldn't otherwise open.

Scope tightly, then run (backgrounded — the harness notifies you):

explore submit --file ~/.davstack/specs/<slug>.md

When the current conversation already contains the real context, prefer compact mode over writing a spec:

explore submit --compact-mode "audit supervisor handoffs"

Keep the inline compact prompt to roughly 5-10 words. Do not paste details, requirements, quotes, or file lists into it; the compact spec-writer reads the current transcript/history and distills that context for the executor. Trust the subagent handoff unless the task is genuinely too ambiguous from conversation history.

For a single scoped fact, skip the spec file — inline it (no boilerplate):

explore submit '<goal>Exact signature + return type of resolve_query_adapter</goal> <scope>backend/src/query/adapter.py only</scope>'

Many --file run in parallel from one command. Read the result → <path> file for the answer.

Begin every --file spec with a markdown # 3-5 word title line — a short overview of the task. The TUI agent viewer renders this as the job label; without it the viewer falls back to the first 5 words of the spec, which is rarely meaningful. (Inline single-fact submits can skip the heading.)

The spec is just goal / context (the one gotcha) / scope tags. Do NOT add an output section — the structured path:line deliverable is automatic.

# Trace report-section autosave path
<goal>How does report-section autosave reach the backend?</goal>
<context>TipTap editor; suspect a debounce plus a websocket path.</context>
<scope>react/src/feat/report/** and any agent ws handler it hits.</scope>

Verify (non-negotiable). Quoted path:line facts are reliable; the synthesis is a hypothesis you re-derive yourself. Tell the user what you delegated and that you verified.

Plus depuis ce dépôt

même dépôt

fix-at-commits

DawidWraga/davstack

Strip the stray leading `@` from unpushed commit messages (a PowerShell here-string glitch that buries the real subject under a bare `@` line). Use when commits show up as `@ <subject>` in git log or the first line looks blank/invisible.

2026-06-1678

fast-edit

DawidWraga/davstack

Large, mechanical, low-complexity edits you can convey in a SHORT intent+constraints spec that the executor will then fully apply on its own — simple refactors, rote multi-file changes, repeated patterns. Worth delegating even if you've already read the files. Do it yourself if 1-2 line edit, or conveying it correctly would need exhaustive line-by-line detail right. Never use critical, complex edits that require careful judgement.

2026-06-0878

diagnose

DawidWraga/davstack

Disciplined diagnosis loop for hard bugs and performance regressions. Frame → reproduce the USER's failure → hypothesis tree → pre-registered experiment → instrument (delegated) → interpret with a confound gate → root cause → fix-as-hypothesis. Backed by a local Sentry log-ingest sink + `diag` query CLI and a Vitest daemon for fast unit/storybook reruns. Use when the user says "diagnose this" / "debug this", reports a bug, says something is broken/throwing/failing, or describes a performance regression.

2026-06-0178

logs-server

DawidWraga/davstack

Query a local Sentry-shaped log sink (per-session SQLite under `.davstack/logs/<name>.db`) for cross-service timelines, error windows, or `trace_id` / `run_id` slices. Use when the user references a failed request, agent run, or test invocation and you need the actual log timeline rather than guessing. Also use when planting hypothesis-driven probes during a `diagnose` loop. Skip for tail-following live stdout (use the dev server directly) or for prod logs (this is local-only).

2026-06-0178

vitest-server

DawidWraga/davstack

Rerun a vitest test or storybook story file against a warm long-lived daemon — story/unit reruns drop from ~50s cold to ~3-15s warm. Use when the user asks to run / rerun a specific `.test.ts` / `.stories.tsx` file, or you've just edited code and want a fast pass/fail cycle. Skip for cold one-off CI checks (`vitest run` is fine) or when configuring vitest itself (daemon caches go stale).

2026-05-2978

name	explore
description	DEFAULT for multi-file codebase exploration ("how does X work / where is Y", tracing an unfamiliar subsystem, log analysis). Delegate the read-only sweep instead of self-grepping — protects your context. Never delegate design/synthesis. Self-check ONLY the most imporant load-bearing files/lines; delegate breadth you wouldn't otherwise open.

Scope tightly, then run (backgrounded — the harness notifies you):

explore submit --file ~/.davstack/specs/<slug>.md

When the current conversation already contains the real context, prefer compact mode over writing a spec:

explore submit --compact-mode "audit supervisor handoffs"

For a single scoped fact, skip the spec file — inline it (no boilerplate):

explore submit '<goal>Exact signature + return type of resolve_query_adapter</goal> <scope>backend/src/query/adapter.py only</scope>'

Many --file run in parallel from one command. Read the result → <path> file for the answer.

The spec is just goal / context (the one gotcha) / scope tags. Do NOT add an output section — the structured path:line deliverable is automatic.

# Trace report-section autosave path
<goal>How does report-section autosave reach the backend?</goal>
<context>TipTap editor; suspect a debounce plus a websocket path.</context>
<scope>react/src/feat/report/** and any agent ws handler it hits.</scope>

Verify (non-negotiable). Quoted path:line facts are reliable; the synthesis is a hypothesis you re-derive yourself. Tell the user what you delegated and that you verified.