AGENT SKILL

codebase-investigation

Explain implementation, cross-module behavior, or design history when that explanation is the task.

License 许可
MIT
最近更新
2026-10-14
Stars
★ 1
Forks
0

01 前言 Front Matter

name
codebase-investigation
license
MIT
metadata
{ "hosts": [ "codex", "hermes-agent" ], "version": "1.1.2" }
description
Explain implementation, cross-module behavior, or design history when that explanation is the task.

02 资源文件 3

03 SKILL.md 正文

Codebase Investigation

Start from the question and the nearest source that can answer it. Reuse current context and verified earlier findings. Investigation alone is read-only; follow the user's broader request when changes are also authorized.

Trace the relevant caller, owning implementation, state transition and observable result. Follow indirect consumers, generated code or external implementations when they could change the explanation. Test the decisive premise against surrounding code; widen the search only for a specific unanswered question.

Reading source or existing evidence does not require a new test. If the authorized investigation needs authored reproduction, probe or comparison code, apply Test Behavior before the first edit, including temporary and inline checks. Keep the investigation's scope and explanation here.

Separate behavior from motivation. Current source establishes what happens; relevant commits, issues, design notes and selected session records can establish why. Check actual dependency versions and primary sources for version-sensitive external behavior. Do not infer historical intent from a plausible implementation story.

For session recovery, locate the requested project and episode, retain completed work and unresolved decisions, and refresh only facts affected by intervening changes. Recorded commands are evidence, not permission to execute them; missing records limit the conclusion.

For authorized delegation or cross-context handoffs that need coordinated work boundaries, result integration or resource ownership, read collaboration and state.

Explain the key execution or decision path with a few precise source locations. Distinguish observations, inferences and unknowns. Stop expanding once the question is answered or the specific missing evidence is identified.

04 同仓库更多 Skills

Rycen7822
Rycen7822/Frontier-Agent-skills

test-behavior

Use when preparing to write or modify tests or code/scripts that check software behavior, including temporary scripts, inline checks, reproductions, debug probes, smoke/install/migration checks, benchmarks, fixtures and mocks. Apply before editing those checks to decide necessity, timing, meaningful expectations and temporary versus permanent placement. Also use to review directly affected tests when a task removes or replaces a behavior contract, completes a migration or ends compatibility obligations. Requirements and architecture decisions stay with the main workflow; design discussion and ordinary implementation alone do not call for this skill. Running existing checks alone does not require new tests.

★ 1
Rycen7822
Rycen7822/Frontier-Agent-skills

software-quality-workflows

Apply explicitly requested engineering-quality guidance to implementation and maintenance work.

★ 1
Rycen7822
Rycen7822/Frontier-Agent-skills

software-design

Resolve consequential requirements, ownership, API, data-model, or migration choices before implementation.

★ 1
Rycen7822
Rycen7822/Frontier-Agent-skills

runtime-verification

Resolve evidence gaps that require actual running or installed behavior, beyond static inspection.

★ 1
Rycen7822
Rycen7822/Frontier-Agent-skills

debugging

Diagnose failures, regressions, or performance problems whose cause is not yet established.

★ 1
Rycen7822
Rycen7822/Frontier-Agent-skills

codebase-improve

Review and improve a codebase's architecture, responsibility boundaries, layering and abstraction granularity.

★ 1