Skip to content

OthmanAdi/planning-with-files / .pi/skills/planning-with-files

planning-with-files

Flagged

Persistent file-based planning for multi-step AI-agent work. 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. Explicit session-catchup.py --metadata reads same-project local agent session records and emits aggregate counts only; --replay may emit bounded nonce-framed excerpts. Optional gated mode can request continuation only when the host supports it and never runs commands declared in Markdown. The skill has no network upload path. Use for research or work needing 5+ tool calls.

FileKindSize
LICENSEother1.0 KB
README.mdmarkdown116.9 KB
SKILL.mdskill md41.0 KB
examples.mdmarkdown4.4 KB
extensions/planning-with-files/README.mdmarkdown2.9 KB
extensions/planning-with-files/__tests__/attestation.test.tsscript8.8 KB
extensions/planning-with-files/__tests__/phase-counts.test.tsscript1.5 KB
extensions/planning-with-files/__tests__/plan-anchor.test.tsscript10.7 KB
extensions/planning-with-files/__tests__/runtime.test.tsscript36.3 KB
extensions/planning-with-files/attestation.tsscript1.3 KB
extensions/planning-with-files/constants.tsscript1.5 KB
extensions/planning-with-files/index.tsscript211 B
extensions/planning-with-files/package-lock.jsonmanifest120.5 KB
extensions/planning-with-files/package.jsonmanifest337 B
extensions/planning-with-files/plan.tsscript13.5 KB
extensions/planning-with-files/runtime.tsscript29.9 KB
package.jsonmanifest1.5 KB
plan-inventory.mdmarkdown9.3 KB
reference.mdmarkdown8.3 KB
scripts/attest-plan.ps1script27.4 KB
scripts/attest-plan.shscript27.4 KB
scripts/check-complete.ps1script12.8 KB
scripts/check-complete.shexecutablescript13.6 KB
scripts/gate-stop.shscript1.4 KB
scripts/init-session.ps1script19.5 KB
scripts/init-session.shexecutablescript15.4 KB
scripts/inject-plan.pyscript66.1 KB
scripts/inject-plan.shscript63.5 KB
scripts/ledger-append.ps1script6.7 KB
scripts/ledger-append.shscript11.9 KB
scripts/ledger-summary.ps1script6.7 KB
scripts/ledger-summary.shscript7.8 KB
scripts/phase-status.ps1script10.2 KB
scripts/phase-status.shscript7.4 KB
scripts/plan-doctor.shscript9.0 KB
scripts/resolve-plan-dir.ps1script11.5 KB
scripts/resolve-plan-dir.shscript15.3 KB
scripts/session-catchup.pyexecutablescript36.4 KB
scripts/set-active-plan.ps1executablescript42.7 KB
scripts/set-active-plan.shexecutablescript38.6 KB
scripts/skill-hook.shexecutablescript19.3 KB
scripts/verify-shell-line-endings.mjsscript1.3 KB
templates/analytics_findings.mdmarkdown1.8 KB
templates/analytics_task_plan.mdmarkdown2.5 KB
templates/findings.mdmarkdown1.1 KB
templates/loop.mdmarkdown2.1 KB
templates/progress.mdmarkdown1.5 KB
templates/task_plan.mdmarkdown2.2 KB

extensions/planning-with-files/README.md

View on GitHub
# planning-with-files Pi Extension

This extension provides lifecycle automation for the `planning-with-files` skill in Pi.

## Events mapped

- `session_start` -> session catchup
- `before_agent_start` -> plan reminder/injection
- `tool_call` -> pre-tool recitation equivalent
- `tool_result` -> post-write reminder
- `agent_end` -> incomplete-task auto-continue (limit 3)
- `session_before_compact` -> compaction reminder and delivery-state invalidation
- `message_end` -> confirm injected planning context
- `session_compact` / `session_tree` -> invalidate stale delivery state

## Modes

- `auto` (default)
- `parity`
- `cache-safe`
- `notify`

### Parity pre-tool reminders

Turn-start injection is unchanged: an approved plan contributes its first 50 lines
and recent progress. Once Pi emits `message_end` for that injection, ordinary
tracked tool calls do not queue the same 30-line pre-tool plan view again, even
as the session leaf advances. A visible edit queues one refresh, including edits
made without a leaf transition. Progress-only edits and edits beyond the first
30 plan lines do not refresh the pre-tool view. Whole-plan attestation is still
checked before suppression; tamper warnings are not suppressed.

Suppression is scoped to the session, selected plan path, and effective mode.
Observed selection/mode changes, missing plans, user input, session start/shutdown,
failed or aborted runs, tree navigation, and compaction invalidate delivery state.
Resume/fork still require `/plan-execute` approval. Compaction invalidates even
when no plan is currently present; canceled compaction may cause an extra reminder.
The `cache-safe` and `notify` modes retain their existing reminder behavior.

To restore the prior per-leaf parity reminder behavior:

```bash
PWF_PRETOOL=always pi
```

This override does not bypass plan selection, execution approval, or attestation.

Delivery bookkeeping is a disposable in-process optimization, not shared planning
state: the authoritative plan remains Markdown on disk, and no persistent cache
or new session metadata is written. Losing bookkeeping permits extra injection.
A queued refresh is distinct from confirmed context and is deduplicated only
while the run remains active; unconfirmed queues are forgotten at `agent_end`.
Pi's `nextTurn` messages enter a fresh run, not the next tool-loop iteration.
Invalidation cannot retract messages already queued by Pi, so recovery can produce
an extra stale reminder before the fresh turn-start plan injection. `message_end`
confirms session-context insertion, not provider receipt or billing. The public
`sendMessage` API exposes synchronous send failures only; these leave no marker.
Context replacement by other extensions (including Observational Memory) without
Pi lifecycle events is not tracked by this patch.

Configure with:

```bash
PWF_MODE=auto pi
```

or in settings (`.pi/settings.json` / `~/.pi/agent/settings.json`):

```json
{
  "planningWithFiles": {
    "mode": "auto"
  }
}
```
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall