audit-todos

Audit active task files against current repository state to classify stale or completed work.

Updated Apr 6, 2026
One-click install
npx skills add https://github.com/toderian/project_template --skill audit-todos-toderian
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: audit-todos
Source: https://github.com/toderian/project_template/tree/main/plugins/agents-tasks/skills/audit-todos
Command: npx skills add https://github.com/toderian/project_template --skill audit-todos-toderian

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Task backlogs drift out of sync with the codebase: tasks get implemented but never closed, become obsolete after refactors, or duplicate each other. This Skill performs an evidence-based audit of active task files against the current repo state so stale backlog work is handled deliberately instead of accumulating. ## Core Features & Use Cases - Evidence-based classification: Each active task is classified as keep, needs-update, appears-done, cancel-or-close, split-follow-up, or needs-user-decision, with cited task and repo evidence. - Structured Markdown report: Produces a grouped report with a summary table, per-task evidence, confidence levels, and a delta section against the previous audit. - Report-only safety: Never edits task files or the worktree; it delegates mutations to existing workflows like complete-task, capture-idea, add-task, and roadmap. - Use Case: Run a periodic backlog health check on docs/tasks_manager/_todos/ to find tasks whose acceptance criteria are already satisfied by current code and tests, then close them through the proper completion workflow. ## Quick Start Ask the agent to audit todos and review all active tasks against the current repository state, then save the report.

Frequently Asked Questions about audit-todos

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit stale tasks in my backlog?

Run the audit over the active tasks directory; it reads each task's metadata, phases, and acceptance criteria, then checks current code, tests, docs, and git history. Every task receives a classification with cited evidence and a recommended next workflow.

How to find tasks that are already implemented in code?

The audit maps each task's acceptance criteria to current code, tests, and docs using file searches and git history. Tasks whose criteria are satisfied are classified as appears-done and routed to the complete-task workflow for verification and closeout.

Does the task audit modify or close tasks automatically?

No, the audit is report-only and leaves the worktree unchanged. It recommends actions but delegates all mutations to separate workflows such as complete-task, capture-idea, add-task, or roadmap, which require explicit user approval.

What happens when no code evidence is found for a task?

Absence of search results is never treated as completion or cancellation. The task is classified as keep if it remains coherent and actionable, or needs-user-decision when evidence is too weak, with the exact searches documented in the report.

Can the audit report track changes between runs?

Yes, when a previous audit report exists, the new report includes a delta section keyed by task ID. It lists tasks entering or leaving each recommendation category and tasks that remained unchanged since the last audit.