file-todos

name	file-todos
description	Structured file-based todo tracking for Rune workflows. Each todo is a markdown file with YAML frontmatter and lifecycle status. Source-aware templates adapt for review findings, work tasks, PR comments, tech debt, and audit findings. Session-scoped: todos live in tmp/{workflow}/{timestamp}/todos/, cleaned by /rune:rest. Use when creating, triaging, resolving, or querying todos from any Rune workflow.
user-invocable	true
disable-model-invocation	false
argument-hint	[create|triage|status|list|next|search|resolve|dedup|manifest] [--status=pending] [--priority=p1] [--source=review]
allowed-tools	["Read","Write","Edit","Bash","Glob","Grep","AskUserQuestion"]

File-Todos — Structured File-Based Todo Tracking

Session-scoped, source-aware todo tracking across all Rune workflows. Each todo is a standalone markdown file with YAML frontmatter for structured metadata and lifecycle tracking. Todos are co-located with their workflow artifacts in tmp/.

Architecture

Directory Structure

Todos are session-scoped — each workflow session owns its own todos/ subdirectory inside tmp/:

Workflow	Todo Base Directory
/rune:strive	tmp/work/{timestamp}/todos/
/rune:appraise	tmp/reviews/{identifier}/todos/
/rune:audit	tmp/audit/{identifier}/todos/
/rune:mend	(uses review/audit session todos — cross-write isolation)
/rune:arc	tmp/arc/{id}/todos/

Within each todos_base, per-source subdirectories organize todos by origin:

tmp/work/1771991022/todos/
├── work/                                     # source: work (from strive)
│   ├── 001-ready-p2-implement-auth-flow.md
│   ├── 002-ready-p3-add-validation.md
│   └── todos-work-manifest.json              # per-source manifest (built by manifest build)
├── review/                                   # source: review (from appraise)
│   ├── 001-pending-p1-fix-sql-injection.md
│   ├── 002-pending-p2-missing-docs.md
│   └── todos-review-manifest.json
├── audit/                                    # source: audit (from audit)
│   ├── 001-pending-p3-dead-code.md
│   └── todos-audit-manifest.json
└── todos-cross-index.json                    # optional cross-source index

ID sequences are per-subdirectory — work/001-* and review/001-* are independent. Sessions with >999 todos use 4-digit IDs (0001-9999).

File Naming Convention

{issue_id}-{status}-{priority}-{slug}.md

Component	Format	Values
issue_id	3-digit padded (4-digit for >999)	001, 002, ... 999, 1000
status	lowercase	pending, ready, complete
priority	lowercase	p1 (critical), p2 (important), p3 (nice-to-have)
slug	kebab-case	Max 40 chars, derived from title

Priority normalization: When reading todo files, normalize priority values from various formats to canonical p1/p2/p3: high/critical -> p1, medium/important -> p2, low/nice-to-have -> p3. This ensures backward compatibility with strive-generated todos that use priority: medium instead of priority: p2.

Tolerant file naming: The reader side accepts BOTH v2 naming (001-ready-p2-slug.md) AND strive-generated naming (task-1.md, task-2.md). The v2 format remains canonical for the create subcommand. Glob patterns throughout this spec include task-[0-9]*.md alongside the standard [0-9][0-9][0-9]-*.md patterns.

Option A (CRITICAL): Filename encodes INITIAL status only. Files are NEVER renamed on status transition. The frontmatter status field is authoritative. The in_progress status exists only in frontmatter, never in the filename.

Slug Algorithm

The slug algorithm MUST be identical everywhere it is used (orchestrator, workers, integrations):

slugify(s):
  1. Convert to lowercase
  2. Replace all non-alphanumeric sequences with a single hyphen
  3. Trim leading/trailing hyphens
  4. Truncate to 40 characters

  Regex: s.toLowerCase().replace(/[^a-z0-9]+/g, '-').replace(/^-+|-+$/g, '').slice(0, 40)

ID Generation

Sequential padded IDs, independent per source subdirectory. Use 3-digit format (001-999); auto-upgrade to 4-digit (0001-9999) for sessions with >999 todos. Use zsh-safe glob for counting existing files:

# zsh-safe: (N) prevents NOMATCH error on empty directory
# source = "work" | "review" | "audit"
# Tolerant: count both v2 naming (NNN-*.md) and strive naming (task-N*.md)
setopt nullglob
existing_3=("${todos_base}/${source}"/[0-9][0-9][0-9]-*.md(N))
existing_4=("${todos_base}/${source}"/[0-9][0-9][0-9][0-9]-*.md(N))
existing_task=("${todos_base}/${source}"/task-[0-9]*.md(N))
total=$(( ${#existing_3[@]} + ${#existing_4[@]} + ${#existing_task[@]} + 1 ))
if (( total > 999 )); then
  next_id=$(printf "%04d" $total)
else
  next_id=$(printf "%03d" $total)
fi

This means work/001-* and review/001-* are independent sequences.

Sole-orchestrator pattern: Only the orchestrator creates todo files. Workers send completion signals; the orchestrator writes. This eliminates TOCTOU race conditions in ID generation.

YAML Frontmatter Schema

Schema v2 adds resolution metadata, ownership audit fields, cross-source linking, and ordering metadata. Schema v1 files remain valid — missing fields default to empty/null.

---
schema_version: 2
status: pending           # pending | ready | in_progress | complete | blocked | wont_fix | interrupted
priority: p1              # p1 (critical) | p2 (important) | p3 (nice-to-have)
issue_id: "001"
source: review            # review | work | pr-comment | tech-debt | audit | test-browser | gap | test
source_ref: ""            # path or reference back to origin
finding_id: ""            # RUNE:FINDING id (for review/audit sources)
finding_severity: ""      # P1/P2/P3 (for display without re-parsing)
tags: []
dependencies: []          # qualified IDs this is blocked by: ["review/001", "work/003"]
files: []                 # affected files (from source)
assigned_to: null         # worker claim tracking
work_session: ""          # DEPRECATED in v2 — retained for backward compat, not populated
created: "YYYY-MM-DD"
updated: "YYYY-MM-DD"

# Resolution metadata (v2)
resolution: null          # null | "fixed" | "false_positive" | "duplicate" | "wont_fix" | "out_of_scope" | "superseded"
resolution_reason: ""     # Free-text explanation (required when resolution is set)
resolved_by: ""           # Agent or user who resolved (e.g., "mend-fixer-1", "user")
resolved_at: ""           # ISO datetime of resolution

# Ownership audit (v2)
claimed_at: ""            # ISO datetime when assigned_to was set
completed_by: ""          # Agent or user who completed (may differ from assigned_to)
completed_at: ""          # ISO datetime of completion

# Cross-source linking (v2)
duplicate_of: ""          # Qualified ID of the original: "{source}/{issue_id}" (when resolution=duplicate)
related_todos: []         # Qualified IDs of related todos: ["review/003", "work/005"]
workflow_chain: []        # Ordered workflow events: ["appraise:1771234", "mend:1771235"]

# Ordering metadata (v2 — computed by manifest build)
execution_order: null     # Topological sort position (null until manifest build runs)
wave: null                # Wave group number (null for manual todos)
---

Status lifecycle (full):

pending --> ready --> in_progress --> complete
  |           |           |             (resolution: fixed)
  |           |           |
  |           |           +--> blocked --> in_progress (when unblocked)
  |           |
  +-----------+---------------> wont_fix
  |                             (resolution: false_positive | duplicate | out_of_scope | superseded | wont_fix)
  |
  +-- in_progress --> interrupted (session ended before completion)
                       --> ready (on session resume)

• pending — Created, awaiting triage
• ready — Triaged and approved for work
• in_progress — Actively being worked on (frontmatter-only, NEVER in filename)
• complete — Work finished, acceptance criteria met (resolution: fixed)
• blocked — Blocked by unresolved dependencies
• wont_fix — Not being addressed; resolution field specifies why
• interrupted — Session ended before completion; transitions back to ready on resume

Note: wont_fix is both a status value AND a resolution outcome. The resolution field provides the specific reason within wont_fix (false_positive, duplicate, out_of_scope, superseded, or wont_fix for risk-accepted).

Configuration

File-todos are always active in all Rune workflows. Todos are mandatory — every workflow produces and consumes file-todos as the foundation for task management and teammate assignment.

v3.x: file-todos defaults are hardcoded (see v3-defaults.md):

• triage.auto_approve_p1: false — P1 findings always require triage approval
• Todos are session-scoped in tmp/{workflow}/{timestamp}/todos/, not at project root

Session-Scoped Design

Todos live inside their workflow's tmp/ directory and are cleaned by /rune:rest. This means:

• No stale todos from abandoned sessions
• Each session owns its todo space (no cross-session interference)
• Todos co-locate with TOME, patches, and other workflow artifacts for traceability
• /rune:file-todos subcommands auto-detect the most recent active session's todos_base

Sub-Command Dispatch

Route sub-commands via $ARGUMENTS. Parse the first argument as the sub-command name. See subcommands.md for full implementation details, validation patterns, and helper functions.

const args = "$ARGUMENTS".trim()
const subcommand = args.split(/\s+/)[0] || ""
const restArgs = args.slice(subcommand.length).trim()

switch (subcommand) {
  case "create":   // -> create workflow
  case "triage":   // -> triage v2 (resolution-aware)
  case "status":   // -> status report
  case "list":     // -> filtered listing
  case "next":     // -> next ready todo
  case "search":   // -> full-text search
  case "resolve":  // -> mark resolution with metadata
  case "dedup":    // -> detect potential duplicates
  case "manifest": // -> manifest subcommand: build | graph | validate
  default:         // -> show usage help
}

For the manifest subcommand, parse the second argument as the manifest subcommand:

case "manifest": {
  const manifestSub = restArgs.split(/\s+/)[0] || ""
  switch (manifestSub) {
    case "build":    // -> build per-source manifests
    case "graph":    // -> dependency graph visualization
    case "validate": // -> validate manifest integrity
    default:         // -> show manifest usage
  }
}

Validation Patterns

All user-supplied filter values MUST be validated before use. Invalid values produce a clear error message, NOT an empty list.

STATUS_PATTERN  = /^(pending|ready|in_progress|complete|blocked|wont_fix|interrupted)$/
PRIORITY_PATTERN = /^p[1-3]$/
SOURCE_PATTERN  = /^(review|work|pr-comment|tech-debt|audit)$/
TAG_PATTERN     = /^[a-zA-Z0-9_-]+$/
TODO_ID_PATTERN = /^[0-9]{3,4}$/

Sub-Commands

create — Interactive Todo Creation

/rune:file-todos create

1. Prompt for title/priority/source/files via AskUserQuestion (source selection determines subdirectory)
2. Ensure source subdirectory exists: mkdir -p {todos_base}/{source}/
3. Generate next sequential ID from source subdirectory (zsh-safe (N) glob, per-subdirectory sequence)
4. Compute slug, write file using todo-template.md
5. Report: "Created {todos_base}/{source}/{filename}"

Zero-state: Auto-creates {todos_base}/{source}/ if missing.

triage — Batch Triage Pending Items (v2)

/rune:file-todos triage

Scan all source subdirectories. Process pending todos sorted by priority (P1 first), capped at 10 per session. See triage-protocol.md for the full v1 workflow; see subcommands.md for the full v2 implementation with resolution-aware options.

v2 options per item: Approve (ready), Defer (keep pending), False Positive (wont_fix + resolution), Duplicate (wont_fix + duplicate_of), Out of Scope (wont_fix + out_of_scope), Superseded (wont_fix + superseded).

In v3.x, P1 auto-approve is OFF (no config layer); every item is presented for triage.

Zero-state: "No pending todos found. All items have been triaged."

status — Summary Report

/rune:file-todos status

Scan all source subdirectories ({todos_base}/*/[0-9][0-9][0-9]-*.md) and display counts grouped by source subdirectory, status, and priority. Output is PLAIN TEXT with no emoji.

File-Todos Status
------------------------------
 Pending:     3 (needs triage)
 Ready:       5 (approved)
 In Progress: 2 (active)
 Complete:    12 (done)
 Blocked:     1
 Wont Fix:    0
------------------------------
 P1:  2 pending, 1 ready
 P2:  1 pending, 3 ready
 P3:  0 pending, 1 ready
------------------------------
 By Source Subdirectory:
   work/       7 (4 complete, 2 ready, 1 in_progress)
   review/     8 (5 complete, 2 pending, 1 blocked)
   audit/      3 (2 pending, 1 complete)
   pr-comment/ 2 (1 complete, 1 pending)
------------------------------

Scan pattern: Tolerant glob matching all naming formats:

Glob(`${todosBase}/*/[0-9][0-9][0-9]-*.md`)
  .concat(Glob(`${todosBase}/*/[0-9][0-9][0-9][0-9]-*.md`))
  .concat(Glob(`${todosBase}/*/task-[0-9]*.md`))

where todosBase is resolved via resolveSessionContext($ARGUMENTS). Only scans the current session's todos_base. Arc todos, work todos, and review todos are all session-scoped and accessible via this mechanism.

Zero-state: "No todos found. Run /rune:file-todos create or run a review/work workflow to auto-generate todos."

list — Filtered Listing

/rune:file-todos list [--status=pending] [--priority=p1] [--source=review] [--tags=security,api]

Scans all source subdirectories using tolerant globs ({todos_base}/*/[0-9][0-9][0-9]-*.md, */[0-9][0-9][0-9][0-9]-*.md, and */task-[0-9]*.md). Filters compose as intersection. The --source filter restricts to a specific subdirectory (e.g., --source=review scans only {todos_base}/review/). Invalid filter values produce a clear error, not an empty list. Sort: priority (P1 first), then issue_id ascending. See subcommands.md for filter parsing and intersection logic.

Zero-state: "No todos match the given filters." or "No todos found."

next — Highest Priority Unblocked Ready Todo

/rune:file-todos next [--auto]

Scans all source subdirectories using tolerant globs (*/[0-9][0-9][0-9]-*.md, */[0-9][0-9][0-9][0-9]-*.md, and */task-[0-9]*.md). Show highest-priority unblocked todo with status: ready and no assigned_to. Checks dependencies against non-complete todos.

--auto flag: Output JSON, claim atomically via lockfile guard (temp-file-then-rename, NOT flock). Sets assigned_to, claimed_at, status: in_progress in frontmatter. See subcommands.md for atomic claim protocol.

Zero-state distinctions:

• No todos at all: "No todos found."
• All complete: "All todos are complete."
• All claimed: "All ready todos are assigned. Check in-progress items."
• None ready: "No ready todos. Run /rune:file-todos triage to approve pending items."
• All blocked: "All ready todos are blocked by unresolved dependencies."

search — Full-Text Search

/rune:file-todos search <query>

Case-insensitive search across all source subdirectories using tolerant globs (*/[0-9][0-9][0-9]-*.md, */[0-9][0-9][0-9][0-9]-*.md, and */task-[0-9]*.md). Searches todo titles, problem statements, and work logs. Validates query length (2-200 chars), sanitizes regex metacharacters before Grep. Results grouped by file with metadata, showing source subdirectory. See subcommands.md for sanitization and display details.

Zero-state: "No matches found for '{query}'."

resolve — Mark Resolution with Metadata

/rune:file-todos resolve 001 --false-positive "Constant value, not user input"
/rune:file-todos resolve 003 --duplicate-of review/001 "Same SQL injection issue"
/rune:file-todos resolve 005 --wont-fix "Legacy code, planned removal in Q2"
/rune:file-todos resolve 007 --out-of-scope "Performance optimization for next sprint"
/rune:file-todos resolve 009 --superseded "Rolled into refactor plan"
/rune:file-todos resolve 003 --undo

Set resolution metadata on a todo. Updates resolution, resolution_reason, resolved_by, resolved_at, and status. --undo reverts a resolution to the previous status. See subcommands.md for full protocol.

Zero-state: "Todo #NNN not found in current session."

dedup — Detect Potential Duplicates

/rune:file-todos dedup
/rune:file-todos dedup --auto-resolve

Scan session todos for potential duplicates using composite scoring (Jaro-Winkler + Jaccard file overlap + finding type + source_ref). Flags pairs with confidence >= 0.70. --auto-resolve auto-marks >= 0.90 confidence pairs as duplicate. See subcommands.md for algorithm details.

Zero-state: "No duplicate candidates found."

manifest build — Build Per-Source Manifests

/rune:file-todos manifest build
/rune:file-todos manifest build --all
/rune:file-todos manifest build --cross-source
/rune:file-todos manifest build --dedup

Build (or incrementally rebuild) todos-{source}-manifest.json files for the current session. Runs Kahn's topological sort, wave assignment, and critical path analysis. See manifest-schema.md for schema and dag-ordering.md for algorithm.

Zero-state: "No source subdirectories found in current session."

manifest graph — Dependency Visualization

/rune:file-todos manifest graph [--source=work] [--all-sources] [--waves] [--mermaid]

Display ASCII dependency graph (default) or Mermaid diagram (--mermaid). Auto-detects primary source for the active workflow. --all-sources requires cross-index (build with manifest build --cross-source). See subcommands.md for output examples.

manifest validate — Validate Integrity

/rune:file-todos manifest validate

Validate per-source manifests for circular dependencies, dangling references, missing resolution reasons, and schema issues. P1 errors are blocking; P2 warnings are advisory. See subcommands.md for validation tier table.

Performance: Per-Source Manifest Cache

For sessions with >100 todos, per-source manifests (todos-{source}-manifest.json) avoid re-parsing all frontmatter. Per-source dirty-signal invalidation: any modifying sub-command writes {source}/.dirty; next manifest build rebuilds only dirty sources via atomic tmp-rename. Manifest is optional — fallback to direct parsing on missing or corrupt manifest. See manifest-schema.md and subcommands.md for full protocol.

Integration Points

Trigger	Flow	Source	Reference
/rune:appraise	TOME findings -> per-finding todos	review	integration-guide.md
/rune:audit	TOME findings -> per-finding todos	audit	integration-guide.md
/rune:strive	Plan tasks -> per-task todos	work	integration-guide.md
/rune:mend	Resolution tracking -> todo updates	updates existing	integration-guide.md
arc Phase 7.7 (TEST)	E2E failures -> per-route todos	test-browser	integration-guide.md
PR comments	gh API -> per-comment todos	pr-comment	integration-guide.md

Concurrent Access

Operation	Who Can Perform	Safe?
Create todo (Write)	Orchestrator only	Yes — single writer
Read todo (Read)	Anyone	Yes — reads are safe
Append Work Log (Edit)	Worker assigned to task only	Yes — one worker per task
Update frontmatter status	Orchestrator or mend-fixer	Conditional — needs claim lock
Delete (cleanup)	/rune:rest cleanup phase only	Yes — after workflow completes

References

• todo-template.md — Source-aware todo template with YAML frontmatter schema (v2)
• lifecycle.md — Status transitions and lifecycle rules (resolution-aware v2)
• triage-protocol.md — Batch triage workflow for pending items
• integration-guide.md — How each Rune workflow feeds/consumes todos
• subcommands.md — Full sub-command implementation details, helpers, session context, and manifest protocol
• manifest-schema.md — Per-source manifest JSON schema and cross-source index format
• dag-ordering.md — Kahn's algorithm, wave assignment, and critical path implementation