TASK GUIDE · SOURCE-AWARE

Agent Skills for codebase and architecture planning

A source-aware Skill plan for mapping a codebase, sequencing work, and recording decisions.

Build this plan for my Agent ↗
Desired outcome

Produce an implementation-ready plan that reflects the repository's real dependencies.

Safe working boundary

Planning does not authorize implementation; unresolved assumptions remain visible.

01 · SKILL STACK

A minimal three-role stack

Each role has one job. The personalized builder can adjust the selection for your Agent without changing the decision logic.

01 · Core execution

improve-codebase-architecture

Maps files, dependencies, tests, and reference patterns.

Main limitation
Architecture quality depends on change history and domain context that may be incomplete.
First risk
Premature refactoring
Source evidence
Pinned · 885e2ca4
Inspect the Skill ↗
02 · Guardrail

repo-intake-and-plan

Breaks the change into ordered, testable phases.

Main limitation
It does not create environments, download assets, execute commands, perform paper lookup, or write the final report.
First risk
Planning from incomplete repository evidence
Source evidence
Pinned · 3ab50525
Inspect the Skill ↗
03 · Evidence

build-evidence-map

Records alternatives, consequences, and the evidence behind the choice.

Main limitation
Validation proves graph integrity, not source quality or inference correctness.
First risk
Sensitive excerpts in portable artifacts
Source evidence
Pinned · 318066d2
Inspect the Skill ↗

02 · ACCEPTANCE

How to know the task is finished

  1. Every phase names its inputs, changed paths, and verification step.
  2. Dependencies and rollback boundaries are explicit.
  3. Rejected alternatives include a concrete reason.

03 · STOP CONDITIONS

When to stop

  • The plan assumes files or APIs that were not inspected.
  • Implementation begins before critical assumptions are resolved.
  • The proposed change has no testable completion condition.

04 · METHOD

How this guide was produced

SkillSignal maps a defined task to three non-overlapping responsibilities, then selects from reviewed catalog metadata. Agent fit, source state, permissions, and published scores are evaluated with deterministic rules. No LLM writes or reorders the recommendation at request time.