Superpowers
obra/superpowersA 15-skill development methodology that makes coding agents clarify the goal, write a plan, work through it with subagents and verify before calling it done.
Recommended
Check first Depth 3/5
Planning with Files Claude skills
Keeps task_plan.md, findings.md and progress.md on disk and re-injects them via hooks, so a plan survives /clear, compaction and crashes.
Our verdict Recommended
A mature, actively maintained planning skill with a real completion gate and recovery path; expect hooks that run on every prompt and tool call.
What Planning with Files does
Planning with Files gives the agent three markdown files in the project, task_plan.md, findings.md and progress.md, and uses lifecycle hooks to put the relevant plan back in front of it. Hooks fire on each prompt and before tool calls such as Write, Edit, Bash, Read, Glob and Grep, so after a /clear or a context compaction the agent picks up from the written plan. A completion gate can hold the agent's stop until the plan reports complete. An explicit session-catchup script can read same-project local agent session records to count or replay what happened, but only when you ask for it; automatic recovery reads project planning files only. The repo is large: 15 skill folders carry the same skill in variants for different agents, IDEs and languages, with a 38 KB main SKILL.md, templates, about 250 script files, a few tests and CI. The description states the skill has no network upload path. It documents native plugins or install routes for Claude Code, Codex, Cursor, OpenCode, Copilot, Gemini and others through npx skills add.
Good for
- Multi-hour tasks that cross several /clear or compaction events
- Teams who want an auditable plan file in the repo
- Anyone on several agents who wants one planning convention
Not for
- Short tasks where a plan file is overhead
- Setups where you do not want hooks running on every tool call
What's inside depth 4/5
A 38 KB main skill plus 14 per-agent variants, about 24 reference files, 75 template files, roughly 250 script files, a few tests and CI, but only two eval files.
- Skills
- 18
- Reference docs
- 26
- Scripts
- 317
- Tests / evals
- 6
- Templates
- 88
- Avg SKILL.md
- 18 KB
Counted from the repository file tree. How we score depth.
Dependencies and costs
- Shell and Python available for the hook and recovery scripts
Security review
Broad accessThis skill asks for more access than most, or sends data to another service. That is usually a design choice rather than bad intent, but you should know before you install it.
- Pre-approves broad shell commandsFrontmatter allowed-tools is "Read Write Edit Bash Glob Grep", which pre-approves unrestricted Bash (the same line is in the per-harness copies).skills/planning-with-files/SKILL.md:5
A file-based planning skill that keeps task_plan.md, findings.md and progress.md on disk and, through hooks declared in its frontmatter, injects plan context on every prompt and tool call and can block the agent from stopping until phases are complete. It pre-approves unrestricted Bash, runs only local shell/Python scripts, and reads local agent session logs only on explicit request. No network code was found in the scripts we searched, but most copies were not read line by line.
Worth knowing
- Relies on a CLI or package we could not reviewInstall docs use
npx skills add OthmanAdi/planning-with-files ... -g,npm install planning-with-files, the /plugin marketplace andhermes plugins install; those installers' code is not in the repo.README.md:55
What we checked
The repository (770 files) on Oct 29, 2026. Not a full review: 313 of 770 files were not downloaded (large repository; we reviewed the skill folders); we read 2 of 319 code files; it relies on code outside the repository: npm packages skills and planning-with-files (installers), and the opencode-planning-with-files npm plugin; 313 readable files in the repo were not available to us (plugin manifests, .codex/hooks scripts, other host adapters); Only a subset of the 319 listed code files was opened: the 75 distinct scripts are repeated across about a dozen per-harness folders, and large files such as inject-plan.py (67 KB) and session-catchup.py were searched for network, process and session-store access rather than read line by line.
NVIDIA SkillSpector static scan: 394 pattern hits (Supply Chain 169, Data Flow 56, MCP Rug Pull 44, Agent Snooping 32, Tool Misuse 31, Dangerous Code Execution 28). The CRITICAL tainted-flow hit is inject-plan.py running its own ledger-summary script on a private temp directory it created; 'Concealed Executable Artifact' hits are the same scripts copied into each harness folder; credential hits are Python's secrets.token_hex for temp file names; 'git push --force' and 'sudo' appear in the Pi extension's blocklist of commands to refuse; session-catchup.py reads local Claude/Codex session transcripts only when the user explicitly asks.
A review, not a certification. Skills change after we check them — read the files before installing.
License
MIT — commercial use allowed.
Can it run unattended?
Yes. The plan lives on disk, is re-injected every turn, and an optional completion gate can keep the agent from stopping early, which suits long unattended runs.
Alternatives to Planning with Files
Continuous Claude
parcadei/Continuous-Claude-v3A personal Claude Code infrastructure: ~160 skills, 32 agents and 30 hooks that inject skill hints, isolate subagent context to files, and fight compaction.
With caveats
Broad access Depth 3/5
Pro Workflow
rohitg00/pro-workflowA Claude Code plugin with 41 skills and many hooks that stores your corrections in SQLite and adds searchable research wikis, quality gates and cost tracking.
With caveats
Check first Depth 2/5
memsearch
zilliztech/memsearchCross-agent persistent memory for Claude Code, Codex, OpenClaw, OpenCode and DeepSeek Harness: Markdown files as the source of truth, with a Milvus search index.
With caveats
Broad access Depth 3/5
FAQ
What is planning-with-files?
It is an agent skill that keeps a task plan, findings and progress log in three markdown files and re-injects them through hooks. The goal is that the plan survives context resets and long-running work.
How do I install planning-with-files?
The README's quick install is npx skills add OthmanAdi/planning-with-files --skill planning-with-files -g. There are also npm and Claude Code plugin marketplace routes, and docs/installation.md lists the rest.
Does planning-with-files run code on my machine?
Yes. It registers command hooks that run a shell script when you submit prompts and before tool calls, and it ships Python and shell helpers. The scripts are readable in the repo, and the README says they do not upload data.
planning-with-files vs Superpowers: which is better?
Planning with Files is narrowly about durable plan state and recovery after context loss. Superpowers is a full process covering design, TDD and review. They can be used together, since one stores the plan and the other shapes the work.
Sources: GitHub repository (stars, license, file tree, last push Oct 28, 2026). Review based on the README and SKILL.md files as published.