Claude Skill

repo-debugging-ledger

Debug repositories with a hypothesis ledger, checked locations, and loop-breaking discipline. Use when tests fail, CI fails, behavior regresses, local tooling errors, flaky bugs appear, or an investigation risks repeated searches, repeated commands, circular hypotheses, or unveri

LLM Mart · 0 points · 1 views 0 listing impressions 0 install-command copies
Download markoblogo-abvx-agent-skills-skills_repo-debugging-ledger-37a7146.zip · 2 KB

Install

skills CLI npx skills add https://github.com/markoblogo/abvx-agent-skills/tree/main/skills/repo-debugging-ledger
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install markoblogo-abvx-agent-skills@llmmart
Git git clone https://github.com/markoblogo/abvx-agent-skills.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole markoblogo/abvx-agent-skills collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Repo Debugging Ledger

Use this skill as a lightweight overlay on normal debugging. Its job is to prevent loops and untested assumptions.

Start With A Repro Signal

Create the smallest useful pass/fail loop:

  • unit/integration test,
  • CLI command,
  • HTTP request,
  • browser script,
  • fixture replay,
  • log assertion,
  • repeated flake trigger.

If no loop exists, state the best available signal and what artifact would make it deterministic.

Hypothesis Ledger

Before editing, write or mentally maintain 3-5 ranked hypotheses:

H1:
Prediction:
Probe:
Result:
Status: open / supported / falsified

Probe one variable at a time. Prefer evidence that distinguishes between hypotheses.

Checked-Location Ledger

Track what has already been inspected:

Checked:
- file/function/command:
  evidence:
  conclusion:

Do not reopen the same file, rerun the same broad search, or repeat the same command unless a phase change makes it newly relevant: new failing signal, new stack trace, changed code, or a narrowed hypothesis.

Search And Instrumentation

  • Use rg and targeted file reads first.
  • Search for exact symbols, error text, config keys, route names, and call sites before broad terms.
  • Tag temporary logs with a unique marker such as [DEBUG-7b3c].
  • Remove temporary instrumentation before finalizing.
  • If a likely area is exhausted, broaden to a different subsystem instead of cycling through the same files.

Fix Discipline

  • Add or identify a regression check before the fix when a reasonable seam exists.
  • Keep the fix scoped to the supported hypothesis.
  • Re-run the original repro signal after the fix.
  • Run the project’s normal checks before finalizing, or explain the blocker.

Final Report

Include:

  • root cause or best-supported cause,
  • files changed,
  • verification run,
  • residual risks or missing test seam.
Files (abvx-agent-skills)
  • agents
    • openai.yaml 213 B
      Loading…
  • SKILL.md 2.2 KB
    ---
    name: repo-debugging-ledger
    description: Debug repositories with a hypothesis ledger, checked locations, and loop-breaking discipline. Use when tests fail, CI fails, behavior regresses, local tooling errors, flaky bugs appear, or an investigation risks repeated searches, repeated commands, circular hypotheses, or unverified fixes.
    license: MIT
    metadata:
      abvx_status: experimental
      abvx_origin: original
    ---
    
    # Repo Debugging Ledger
    
    Use this skill as a lightweight overlay on normal debugging. Its job is to prevent loops and untested assumptions.
    
    ## Start With A Repro Signal
    
    Create the smallest useful pass/fail loop:
    
    - unit/integration test,
    - CLI command,
    - HTTP request,
    - browser script,
    - fixture replay,
    - log assertion,
    - repeated flake trigger.
    
    If no loop exists, state the best available signal and what artifact would make it deterministic.
    
    ## Hypothesis Ledger
    
    Before editing, write or mentally maintain 3-5 ranked hypotheses:
    
    ```text
    H1:
    Prediction:
    Probe:
    Result:
    Status: open / supported / falsified
    ```
    
    Probe one variable at a time. Prefer evidence that distinguishes between hypotheses.
    
    ## Checked-Location Ledger
    
    Track what has already been inspected:
    
    ```text
    Checked:
    - file/function/command:
      evidence:
      conclusion:
    ```
    
    Do not reopen the same file, rerun the same broad search, or repeat the same command unless a phase change makes it newly relevant: new failing signal, new stack trace, changed code, or a narrowed hypothesis.
    
    ## Search And Instrumentation
    
    - Use `rg` and targeted file reads first.
    - Search for exact symbols, error text, config keys, route names, and call sites before broad terms.
    - Tag temporary logs with a unique marker such as `[DEBUG-7b3c]`.
    - Remove temporary instrumentation before finalizing.
    - If a likely area is exhausted, broaden to a different subsystem instead of cycling through the same files.
    
    ## Fix Discipline
    
    - Add or identify a regression check before the fix when a reasonable seam exists.
    - Keep the fix scoped to the supported hypothesis.
    - Re-run the original repro signal after the fix.
    - Run the project’s normal checks before finalizing, or explain the blocker.
    
    ## Final Report
    
    Include:
    
    - root cause or best-supported cause,
    - files changed,
    - verification run,
    - residual risks or missing test seam.
    
  • SKILL_CARD.md 1.2 KB
    Loading…

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.