mcpbeat

Codebase Intel

wednesday-solutions/codebase-intel

Unified codebase intelligence. Handles all questions about structure, logic, risk, and dependencies. Combines natural-language Q&A with deterministic lookups and pre-edit blast radius checks.

985 tokens
context cost
the whole folder, loaded on every use
1
files
instructions only
0
copies elsewhere
how many repositories repackaged it
164
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/wednesday-solutions/ai-agent-skills --skill codebase-intel

The instruction itself

11 sections, as written by the author

Codebase Intelligence Specialist

This skill provides a comprehensive understanding of the project's structure, intent, and risk. Use it for everything from high-level architecture questions to detailed impact analysis before editing code.

When to use

1. Codebase Discovery & Q&A

  • "What does tokenService do?" (Summaries)
  • "Where is the payment logic located?" (Role/Path mapping)
  • "Show me an overview of the architecture." (Stats & Entry points)
  • "Who last touched this file?" (Git history)
  • "Find circular dependencies or dead code." (Structural audit)

2. Risk Assessment (Before Editing)

  • "Is it safe to change this function signature?"
  • "What is the blast radius of this file?"
  • "What will break if I delete this constant?"

3. Graph Maintenance

  • If a query returns "not mapped" or coverage is low.
  • If you notice missing dependencies in the graph.

How to use — by task type

🔍 Discovery & Q&A

  • File Summary: Bash(wednesday-skills query getFileSummary <file_path>)
  • Returns: Role, Summary, Risk Score, and Blast Radius.
  • Architecture Stats: Bash(wednesday-skills query getCodebaseStats)
  • Use getHighConfidenceEntryPoints to identify the best starting files.
  • Advanced Lookups:
  • Bash(wednesday-skills query getHighRiskFiles 70) — find critical technical debt.
  • Bash(wednesday-skills query getCircularDependencies) — find architectural smells.
  • Bash(wednesday-skills query getAllDeadCode) — find unreachable modules.
  • Context: Read .wednesday/codebase/MASTER.md for danger zones and primary data flows.
  • History: Bash(git log --follow --oneline -20 -- <file>) for authorship.

⚠️ Pre-Edit Safety Check (Mandatory)

Before modifying any file, you MUST perform these checks:

  • Check Risk: Bash(wednesday-skills query getFileSummary <file_path>)
  • Score 0–30: Proceed directly.
  • Score 31–60: Inform dev of the risk, proceed with care.
  • Score 61–80: List direct dependents and transitive count; ask confirmation.
  • Score 81–100: STOP. Require explicit dev approval before touching.
  • Blast Radius: Bash(wednesday-skills blast <file_path>::<symbol_optional>)
  • Review direct/transitive callers. Use this for cross-language impact (Go/Py/JS).

🛠 Graph Maintenance & Gaps

If you hit "not mapped" or detect a missing link:

  • Gap Check: Bash(sqlite3 .wednesday/graph.db "SELECT file_path, meta FROM nodes WHERE file_path LIKE '%<file>%'")
  • Check meta for gaps.eventEmitter, gaps.dynamic, etc.
  • Fill Gaps: Bash(wednesday-skills fill-gaps --file <file> --min-risk 50)
  • *Rule*: Only edges with confidence > 0.70 are added automatically.
  • Annotations: If gaps persist, ask dev to add // @wednesday-skills:connects-to <symbol> → <file>.
  • Refresh: Bash(wednesday-skills analyze --incremental) after adding annotations.

🚫 Never

  • Guess: If data is missing, report "Not mapped" and suggest wednesday-skills map --full.
  • Skip Checks: Never edit a file with risk > 80 without explicit dev confirmation.
  • Token Bloat: Do NOT read raw source files to answer structural questions.
  • Add Unreliable Edges: Never manually add edges with confidence below 0.70.

📄 Source Citation

Always end with the source:

  • graph.db — Structural/Summary data
  • MASTER.md — Architectural context
  • git log — History/Authorship

How to use it

Copy the folder

Take wednesday-solutions/codebase-intel from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.