planning-with-files

Never start a complex task without task_plan.md. Non-negotiable.

"After every 2 view/browser/search operations, IMMEDIATELY save key findings to text files."

This prevents visual/multimodal information from being lost.

Before major decisions, read the plan file. This keeps goals in your attention window.

After completing any phase:

• Mark phase status: in_progress → complete
• Log any errors encountered
• Note files created/modified

Whenever a phase status changes, also refresh ## Next Step in task_plan.md so it names the single next action.

Every error goes in the plan file. This builds knowledge and prevents repetition.

## Errors Encountered
| Error | Attempt | Resolution |
|-------|---------|------------|
| FileNotFoundError | 1 | Created default config |
| API timeout | 2 | Added retry logic |

if action_failed:
    next_action != same_action

Track what you tried. Mutate the approach.

When all phases are done but the user requests additional work:

• Add new phases to task_plan.md (e.g., Phase 6, Phase 7)
• Log a new session entry in progress.md
• Continue the planning workflow as normal

To find a task before resuming it, run sh "<skill-dir>/scripts/set-active-plan.sh" --list or, in Windows PowerShell, & "<skill-dir>/scripts/set-active-plan.ps1" -List. Replace <skill-dir> with this installed skill directory and keep your current directory at the project root.

This read-only command lists named plans and phase progress under the current directory's .planning/. [active] marks the shared default pointer; it does not bind a session. Concurrent tasks still require each host's PLAN_ID or separate worktrees.

For independent tasks in the same repository, create a named plan for each and pin each agent host to its own plan:

# Terminal A: initialize, then use the exact PLAN_ID printed by the script.
./scripts/init-session.sh "Backend Refactor"
export PLAN_ID=2026-09-05-backend-refactor
# Start the agent from this terminal after setting PLAN_ID.

# Terminal B: use the different PLAN_ID printed for this task.
./scripts/init-session.sh "Incident Investigation"
export PLAN_ID=2026-09-05-incident-investigation
# Start the second agent from this terminal.

The IDs above are examples; initialization uses today's date and may add a numeric suffix. In PowerShell, set $env:PLAN_ID to the printed ID before starting the agent. Setting an environment variable inside an already-running agent's tool subprocess does not change the parent host's hook environment. Use separate worktrees when the host cannot be pinned per task.

set-active-plan.sh changes the repository's shared default pointer, so use it for sequential switching. It does not bind concurrent sessions. PWF_PLAN_ROOT chooses a project root; add PLAN_ID when that root contains several tasks. An .attached marker authorizes a session to receive context but does not select its plan. When session isolation is armed and multiple plans exist, the Codex, Hermes, Pi, and standalone hook routes refuse unpinned selection instead of following another session's pointer.

For several agents collaborating on one task, share its PLAN_ID, keep one orchestrator as the plan owner, and give workers separate ledgers or files.

PLAN_ID is a slug resolved against the current directory, so it can only ever name a plan under $(pwd)/.planning. When an agent thread runs with its cwd at a shared parent (/workspace) while the real work lives in a nested project (/workspace/project), the parent's plan is the only one the hooks can see, and it used to be injected on every fire. PWF_PLAN_ROOT takes an absolute path and pins resolution to that root regardless of where the cwd sits. A pin that does not resolve stops injection rather than falling back.

When no pin is set, the plan was picked by the .active_plan pointer or by the newest plan directory, and a project directly below the root carries its own planning state, the hooks treat that as ambiguous and inject nothing:

[planning-with-files] Ambiguous plan: this cwd has an active plan and a nested
project below it has its own (project). Nothing injected. Pin the thread with
PWF_PLAN_ROOT=<absolute path> or PLAN_ID=<slug>.

An explicit PLAN_ID or PWF_PLAN_ROOT can skip that nested-root check. An attachment marker alone cannot. When isolation is armed, several tasks within one root still require PLAN_ID. Detection looks one directory deep, so a project nested further down is not detected.

• scripts/session-catchup.py: With explicit --metadata or --replay, reads same-project records from the active host store. OpenCode uses the read-only SQLite store at ${XDG_DATA_HOME:-~/.local/share}/opencode/opencode.db.

Not every install path ships every surface in this section. Two distinct install routes exist:

Plugin installs register six lifecycle events from hooks/hooks.json, including quiet SessionStart recovery. Standalone skill installs register the five hooks in this SKILL.md frontmatter only after the skill is invoked for that session, so they have no startup recovery. The /plan-goal and /plan-loop slash commands live in commands/ at the repository root and are available from the versioned plugin cache. Skill-only installs land at ~/.claude/skills/planning-with-files/ and do not include commands/.

The standalone scripts/skill-hook.sh reads the host's JSON session identity. UserPromptSubmit emits plain context; PreToolUse and PostToolUse emit the event's additionalContext JSON. The progress reminder fires at most once per turn when a usable session identity and private cache are available, and repeats when those are unavailable. All five events follow the same plan selection and opt-out checks.

Both slash commands carry disable-model-invocation: true, so invoke them explicitly. If a command is unavailable on a skill-only install, the manual fallback below produces the same planning-file result.

Both supported routes register a PreCompact hook with matcher "*". It fires for manual and automatic compaction after the relevant hook route is active. With a selected plan, it prints a diagnostic reminder and the recorded Plan-SHA256 when present. It stays silent without a plan and never blocks compaction.

Claude Code does not support additionalContext for PreCompact. Successful stdout from this event is diagnostic output, so the hook cannot make the model flush progress before compaction. Keep progress current during the task and recover from the selected files on the next prompt. The recorded digest can be compared with the plan bytes; it does not establish human approval.

Composes with Claude Code's /goal. Derives a goal condition from the active plan and forwards it to /goal, so the agent keeps working until the plan file actually reports complete.

/plan-goal                                # default: "all phases report Status: complete"
/plan-goal until all tests pass           # appends user clause to default

/plan-goal does not replace /goal. /goal "anything" still works.

Composes with Claude Code's /loop. Default 10-minute tick re-reads the planning files, runs check-complete, and writes a progress.md entry if nothing changed since the last tick.

/plan-loop                                # default 10m cadence, default tick prompt
/plan-loop 5m                             # override interval
/plan-loop 15m custom prompt              # override interval + prompt

For a "babysit until done" workflow, combine /plan-loop (cadence) with /plan-goal (termination criterion).

For skill-only installs (no commands/ folder) or sessions where the slash command refuses to fire, the model can produce the same effect by executing the wrapper steps inline.

Manual /plan-goal procedure:

1. Resolve the active plan: prefer ${PLAN_ID} env var, then .planning/.active_plan, then newest .planning/<dir>/, then legacy ./task_plan.md.
2. Read the resolved task_plan.md.
3. Compose a goal condition. Default: "all phases in task_plan.md report Status: complete and check-complete.sh reports ALL PHASES COMPLETE". If the user passed additional clauses, append them.
4. Issue Claude Code's native /goal <condition> (CC primitive, always available).
5. Confirm to the user: print the condition + active plan ID + remind that /goal clear cancels.
6. Refuse if task_plan.md does not exist; direct the user to run init first.

Manual /plan-loop procedure:

1. Parse args: first arg matching ^\d+[smhd]$ is the interval (default 10m), remaining args are an optional task prompt.
2. Resolve the active plan as above.
3. Compose the loop tick prompt. If user passed a task prompt, use it verbatim. Otherwise use the planning-aware default that re-reads task_plan.md and progress.md, runs scripts/check-complete.sh, and writes a progress.md entry if no progress was logged since the last tick.
4. Issue Claude Code's native /loop <interval> <prompt> (CC primitive, always available).
5. Confirm to the user: print interval + active plan ID + remind that bare /loop runs the built-in maintenance prompt.

Both procedures match what the commands/plan-goal.md and commands/plan-loop.md files would have fed the model when invoked. The native /loop and /goal primitives are always available in Claude Code; only the planning-aware wrapper is plugin-scoped.

Claude Code's bare /loop reads .claude/loop.md (project) or ~/.claude/loop.md (user). v2.38 ships a planning-aware template at templates/loop.md. Install once:

# Resolve the host-provided installation folder, or set it explicitly.
PWF_SKILL_DIR="${CLAUDE_SKILL_DIR:-${CLAUDE_PLUGIN_ROOT:-$HOME/.claude/skills/planning-with-files}}"
# user-wide
cp "${PWF_SKILL_DIR}/templates/loop.md" ~/.claude/loop.md

# project-specific
cp "${PWF_SKILL_DIR}/templates/loop.md" .claude/loop.md

After install, bare /loop <interval> runs the planning-aware tick.

With no .mode file and no other v3 marker, the hooks produce byte-identical output to v2.43, including the raw progress.md tail and the ===BEGIN PLAN DATA=== / ===END PLAN DATA=== delimiters. Every v3 behavior is additive and opt-in. No existing workflow changes.

Autonomous mode answers the recitation question: strong models drift less, so the per-tool-call plan re-injection (about 90 tokens per matched tool call, the component that scales with tool use) is dropped. Turn-start injection stays because the evidence (arxiv 2603.03258, claudefa.st on Opus 4.7+ subagents) shows drift is real and the full plan file still matters once per turn. Eliminating recitation entirely is not supported by evidence.

Gated mode adds the completion gate on top of autonomous behavior. The gate is the termination oracle: it judges the plan artifact on disk, not the conversation transcript, which is why it beats a transcript-bound evaluator that can be hallucinated.

The default injection is head -50 (turn start) and head -30 (per tool call), which is position-blind: late in a long plan the in_progress phase, the Decisions journal, and the Errors table all sit past the injected window, so every injection pays the token cost while the window no longer carries the active phase. Opt in with PWF_INJECT=smart in the environment, or an inject-smart token in the plan's .mode file, and the injection instead emits: the plan title, the Goal / Next Step / Current Phase sections, a phase count, the full first in_progress phase section, and the last 3 rows of Decisions Made. Plans without ### Phase headings fall back to the plain head. inject-smart alone does not activate any other v3 behavior; it composes with autonomous and gated modes (init-session mode tokens are space-separated in .mode). With neither the env var nor the token present, output is byte-identical to the legacy shape.

Two sessions sharing one plan directory can both write task_plan.md from the same read. The later write silently discards the earlier one's work, and nothing notices: injection, plan-doctor and the Stop gate all read the clobbered file as an ordinary edit. Attestation does not cover this. It compares against a baseline a human approved once, it reports a collaborator's edit with the same [PLAN TAMPERED] wording as a hostile rewrite, and it is a read-side gate that cannot stop the stale write from landing.

The guard compares progress between turn-start fires rather than hashes. Checked items and completed phases only go up during normal work, so a DECREASE means work that was on disk is gone. Forward motion stays silent, which is what keeps the signal worth reading, and both markers are language-neutral because every translated template keeps the literal English **Status:** complete token. On a decrease it prints one advisory line naming how much was lost and pointing at git diff, then injects normally. It never blocks: this hook always exits 0 and this guard does not intercept writes. Archiving completed phases also trips it. Turn it off with PWF_PLAN_GUARD=0 or a plan-guard-off token in .mode.

This is an advisory check after a write, not a lock or merge mechanism. It does not detect overwritten progress.md or findings.md, or plan changes that preserve the completion counts. Keep a single writer for shared summaries and separate files for workers.

Known ceiling: the marker is keyed on the plan path, not the session, so the warning reaches whichever session fires next rather than specifically the one holding the stale copy. Per-session keying needs PWF_SESSION_ID, which most hosts never set.

The Stop gate blocks ONLY when all of these hold. Any single failure allows the stop. This is the lesson from issue #178: an incomplete plan is a normal state, not an error, and accidental blocking infuriates users.

1. Mode is gated (the .mode file contains gate).
2. An in_progress phase exists (not merely COMPLETE < TOTAL).
3. stop_hook_active is false on the Stop hook stdin (already inside a forced continuation means allow stop).
4. Block count is below the cap (default 20, PWF_GATE_CAP to override, reset at init-session).
5. The ledger progressed since the previous block (a stall means allow stop).

The block reason is a fixed template plus the phase NAME only. Plan body text never enters the reason. Outside gated mode the wording is always advisory, never imperative (PR #180 lesson: imperative text in a reason field becomes a continuation command).

The gate mechanism is host-aware. Not every host can hard-block a stop.

Hosts without a blocking Stop hook still get autonomous mode (low recitation + ledger). They do not get gate enforcement; the gate degrades to a notification. This is documented honestly: the gate is real enforcement only on Tier 1.

The gate carries its own guards so a runaway loop cannot run unbounded, independent of any undocumented host behavior:

• Persistent block counter in .planning/<id>/.stop_blocks, reset at init-session. Without the reset, a previous run's count would let the next run stop instantly.
• Cap (default 20) on consecutive blocks. At the cap, the gate allows the stop.
• Stall detection: no new ledger line since the previous block means the model is not progressing, so the gate allows the stop.
• stop_hook_active and the host block cap are backstops, not the primary guard. The counter and stall detector are deterministic and do not depend on undocumented platform fields.

In autonomous and gated mode the raw progress.md tail injection is replaced by a synthesized summary from scripts/ledger-summary.sh. The summary reports tick count, phase complete/total, the in_progress phase heading, and the last event type per agent. No free text from disk reaches the model context, and the block carries no timestamps, so it is KV-cache stable by construction.

The machine ledger lives at .planning/<id>/ledger-<agent>.jsonl, append-only, one JSON object per line. Workers append to their own ledger; the orchestrator owns task_plan.md. The gate's stall detector reads the ledger (a semantic signal) rather than progress.md mtime (which moves on any touch). See scripts/ledger-append.sh and scripts/ledger-summary.sh.

• The skill reads and writes task_plan.md, findings.md, progress.md, and optional .planning/ state in the current project.
• Activated hooks place selected project planning data into model context. External material copied into planning files remains untrusted.
• Automatic recovery and bare session-catchup.py do not inspect host session stores. Explicit --metadata reads same-project local session records and emits aggregate counts only; explicit --replay may emit bounded nonce-framed excerpts.
• The shipped catchup path contains no network request or upload operation. Hook output may still become part of a request made by the host agent to its configured model provider.
• Default Stop behavior is advisory. Optional gated mode can request continuation only through a capable host. It evaluates mode, phase status, Stop-hook state, block count, and ledger progress; it never executes commands declared in Markdown.

1. Delimiter framing (v2.36.1). Plan content is wrapped in BEGIN/END markers and tagged as data. Reduces the surface but does not eliminate prompt injection: the model still parses the content.
2. Hash attestation (v2.37.0; opt-in in legacy mode, default-on in v3 modes). Run /plan-attest (or sh scripts/attest-plan.sh) once you have approved the current plan. The hooks compute a SHA-256 of task_plan.md on every fire and compare against the stored hash. On mismatch, injection is blocked with a [PLAN TAMPERED] warning. This detects a plan-only change while the saved digest remains trusted. The digest is an ordinary local SHA-256 value, not a keyed signature: a process that can replace both the plan and the attestation can make new content pass. Auto-attestation during initialization records the generated bytes; it is not proof of human review. Attestation does not make embedded instructions trustworthy or eliminate model-level prompt injection.

The attestation is written to .planning/<active-plan>/.attestation (parallel-plan mode) or ./.plan-attestation (legacy mode). When set, the injected context also carries a Plan-SHA256: line so the model can log the attested hash for audit.

For the attest-plan.sh write path, optional flock guard, macOS and Windows Git Bash fallback, and why slug-mode is preferred for parallel sessions, see attestation locking and fallback. For the transient SHA cache (location, keying, container behavior, and how to clear it), see performance notes.

These changes apply only when a plan opts into a v3 mode. Legacy plans are unaffected.

• Nonce delimiters. When a plan has a .nonce file (generated at init in v3 modes), the injection wraps plan content in ===BEGIN-PLAN-DATA-<nonce>=== / ===END-PLAN-DATA-<nonce>=== instead of the static markers. A static delimiter inside plan content can break the framing (delimiter-confusion injection); a per-session nonce raises the bar because the delimiter is not a fixed string. The honest limitation: .nonce and task_plan.md live in the same plan directory, so an attacker who can already write task_plan.md can also read .nonce and forge the matching END delimiter. Nonce framing is not an access-control boundary. Attestation detects a plan change only when the attacker cannot also replace the saved digest. In legacy unattested mode, delimiter-confusion injection remains possible for anyone who can write the plan file, so do not rely on the framing alone for prompt-injection defense there. Plans without a .nonce keep the v2 static delimiters.
• Attested injection refusal (v3 modes). Because the nonce cannot defend against an attacker who can write the plan, autonomous and gated mode refuse to inject the plan body at all when no attestation is present: the hook emits [planning-with-files] v3 mode requires attested plan; run attest-plan instead of the plan content. Combined with attestation default-on at init, this means an unattended v3 loop never injects a body without a matching recorded digest. Legacy mode is unchanged: it injects with the v2 static delimiters and attestation stays opt-in.
• Structured ledger injection. In autonomous and gated mode the raw progress.md tail is no longer injected. progress.md is not covered by attestation, so any instruction-like text written there (for example a tool output or a fetched page summary appended during an unattended run) used to flow into context every turn. v3 injects a synthesized ledger-summary.sh block with no free text from disk instead.
• Attestation default-on. Autonomous and gated mode attest the plan at init. Unattended loops amplify any single injection on every tick, so the tamper gate is on from the start, not opt-in. Editing the plan after init requires explicit re-attest.
• User-private SHA cache. The hook SHA cache moved from a world-writable /tmp path to $XDG_CACHE_HOME/pwf-sha (or ~/.cache/pwf-sha), which removes the shared-tmp poisoning surface. In gated mode the cache is a perf hint only: the gate path always re-hashes so the termination oracle never trusts a stale entry.