Conciseness

The body is mostly lean tables and rules, but there is real duplication and padding: the 'planning files go in your project, not the skill directory' point appears three times (the location table, the Quick Start note, and the Anti-Patterns row); 'The Core Pattern' section ("Context Window = RAM (volatile, limited) / Filesystem = Disk") explains a concept Claude already knows; and the Read vs Write Decision Matrix, 5-Question Reboot Test, and Critical Rules substantially restate one another. This fits anchor 3 ('mostly efficient but includes some unnecessary explanation or could be tightened') — above anchor 2 since nothing is tutorial-style filler, below anchor 4 because three sections could be merged or cut outright.

Actionability

Concrete, executable guidance dominates: a copy-paste bash one-liner for session catchup ("$(command -v python3 || command -v python) ~/.codex/skills/planning-with-files/scripts/session-catchup.py --metadata \"$(pwd)\""), a PowerShell equivalent, "run `scripts/init-session.sh \"Task Name\"`", and a fully specified list command ("sh \"<skill-dir>/scripts/set-active-plan.sh\" --list"). Minor gaps keep it at anchor 4 rather than 5: `resolve-plan-dir.sh` is invoked by name with PLAN_ID/PWF_PLAN_ROOT but no usage example is shown, and `check-complete.sh` is listed without the command line or expected output.

Workflow Clarity

The Quick Start is a clear numbered sequence ("Resolve or initialize the task directory" → "Create missing planning files only" → "Re-read the selected plan before decisions" → "Assign one plan owner") with recovery state restored first ("FIRST: Restore Project State"), and error feedback loops exist (the 3-Strike protocol with "AFTER 3 FAILURES: Escalate to User", plus 'Log ALL Errors'). It matches anchor 4 ('clear sequence with most checkpoints present; minor validation gaps') — the missing piece is an explicit completion checkpoint in the sequence, e.g. running `check-complete.sh` to verify all phases are complete before wrapping up.

Progressive Disclosure

The body is a well-sectioned overview with one-level-deep, clearly signaled references that exist in the bundle: "[references/reference.md](references/reference.md)" and "[references/examples.md](references/examples.md)" under Advanced Topics, plus a Scripts section describing each script. It sits at anchor 4 rather than 5 because the three template links ("[templates/task_plan.md](templates/task_plan.md)", etc.) point at a `templates/` directory that is not present in this bundle (the body says templates live at ~/.codex/skills/planning-with-files/templates/), so those links break in-place and force the reader to hunt for the install path.

Specificity

The description lists several concrete mechanisms — "Keeps task_plan.md, findings.md, and progress.md on disk; lifecycle hooks inject selected project planning context", "Automatic recovery reads project planning files only", "session-catchup.py --metadata reads same-project local agent session records and emits aggregate counts only" — matching anchor 4's 'several specific actions; minor gaps in coverage'. It falls short of anchor 5 because a sizable share of the text is implementation/security constraints ("never runs commands declared in Markdown", "no network upload path", "nonce-framed excerpts") rather than user-facing capabilities, and the core actions (create plan, update progress, log findings) are implied rather than enumerated.

Completeness

Both parts are present: the 'what' is clear ("Persistent file-based planning... Keeps task_plan.md, findings.md, and progress.md on disk") and the 'when' is explicit ("Use for research or work needing 5+ tool calls"). It sits at anchor 4 rather than 5 because the 'when' clause is a single terse trigger; the richer trigger set the body itself uses ("Multi-step tasks (3+ steps)", "Building/creating projects", "Tasks spanning many tool calls") is not surfaced in the description, so the 'when' could be more explicit and specific.

Trigger Term Quality

Natural user-facing terms are present: "multi-step AI-agent work", "planning", "research", "work needing 5+ tool calls". This matches anchor 4 ('good keyword coverage; a few natural terms missing') but not anchor 5, since common synonyms users would actually say — "long-running tasks", "project management", "todo", "organize", "complex task" — are absent, and the file-name tokens (task_plan.md) are jargon users rarely say unprompted.

Distinctiveness Conflict Risk

The concrete artifacts ("task_plan.md, findings.md, progress.md", "session-catchup.py", "lifecycle hooks") carve out a recognizable niche, matching anchor 4's 'mostly distinct; minor overlap risk'. Not anchor 5: the broad framing "planning for multi-step AI-agent work" overlaps with generic task-management, todo, and planning skills, and the trigger "research or work needing 5+ tool calls" could plausibly fire for other long-running-workflow skills.