Agent skill

Project Manager

by Th0rgal in Th0rgal/sandboxed.sh

Steer sandboxed.sh projects, roadmaps, decisions from chat. An agent skill from Th0rgal/sandboxed.sh.

MITAuto-check passedProduct & Project Management

Install Project Manager

skills CLI
$ npx skills add Th0rgal/sandboxed.sh --skill project-manager -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Th0rgal/sandboxed.sh project-manager --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Th0rgal/sandboxed.sh.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/project-manager .claude/skills/project-manager && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
project-manager
GitHub stars
515
Token cost
~1.8k tokens
SKILL.md length
938 words
Files
1
Skills in repo
9
Repo updated
First seen
Licence
MIT

At a glance

Steer sandboxed.sh projects, roadmaps, decisions from chat. An agent skill from Th0rgal/sandboxed.sh.

  • Works in 3 steps: list_projects — the roster with buckets… → get_project — objective, status/mode,… → get_project_tasks — the roadmap: board…
  • Tasks that involve Project management
  • SKILL.md covers When to Use, Reading state (do this before…, Shaping the roadmap from… and Decisions and autonomy, plus 1 more section
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Project Manager is an agent skill from Th0rgal/sandboxed.sh. Steer sandboxed.sh projects, roadmaps, decisions from chat.

Its SKILL.md is about 1.8k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Product & Project Management, covering Project management. It works with Model Context Protocol. The repository describes itself as: Safe runtime for autonomous on-chain AI agents: isolated sandboxes, Library skills, encrypted secrets. The licence is MIT.

When your agent uses it

  • Tasks that involve Project management

Example prompts

  • “/project-manager”

Workflow steps

3 steps, taken from the first numbered list in SKILL.md.

  1. list_projects — the roster with buckets (attention/active/paused) and
  2. get_project — objective, status/mode, blocker, next action, grant,
  3. get_project_tasks — the roadmap: board tasks planned by the

What it can do on your machine

Read from SKILL.md and the folder at commit f7dcb98. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Project Manager loads about 1.8k tokens when it runs. Until then it costs about 19 tokens; SKILL.md has 938 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~19
When it runs · the whole SKILL.md, loaded when a task matches
~1.8k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Th0rgal/sandboxed.sh at commit f7dcb98, republished under its MIT licence (© Th0rgal). 938 words, ~1,812 tokens.

Download SKILL.mdSave it as .claude/skills/project-manager/SKILL.md (or your agent's skills folder).
name
project-manager
description
Steer sandboxed.sh projects, roadmaps, decisions from chat.
version
2.0.0
author
thomas
license
MIT
platforms
linux, macos, windows
prerequisites.tools
mcp

Project Manager — projects, roadmaps, decisions

To withdraw work still queued, call cancel_action(action_id, idempotency_key). Read its receipt: cancelled=false means dispatch already started or settled; use the existing mission/job cancellation tool after inspecting the target. Cancelling an MCP request or closing stdio does not cancel accepted work.

When to Use

When the owner asks about their projects ("how is X going?", "what's on the roadmap?", "what needs me?"), wants roadmap items added/edited/cancelled from conversation, answers an escalated decision, or asks to change a project's autonomy grant. Requires the sandboxed_assistant MCP server.

Use the sandboxed_assistant MCP tool family to talk about the owner's projects with real data, never from memory. Every project is identified by its slug.

The server runs sandboxed-mcp with a scoped coordinator credential. Read get_capabilities if a tool is unavailable; only an operator session can change autonomy grants or infrastructure. Do not bypass the boundary with an owner token or a direct API call.

Every mutation requires an idempotency_key and returns an action receipt. Keep the same key and arguments after a lost response. Read get_action until it settles before reporting the change as applied. An action marked reconciliation_required has an uncertain outcome: inspect existing state, do not repeat it under a new key. A completed launch action means its mission was created, not that the mission finished or its evidence was accepted.

For Antigravity attempts, preserve backend="antigravity" and the selected machine's native model ID. An indeterminate native-login status is not evidence of missing credentials. Require target-account discovery and execution receipts; do not substitute the Gemini API backend or another machine's model catalog.

Reading state (do this before opining)

  1. list_projects — the roster with buckets (attention/active/paused) and health. Start here when the owner asks "how are my projects doing?".
  2. get_project <slug> — objective, status/mode, blocker, next action, grant, open decisions, recent decisions (the decision ledger), tracks.
  3. get_project_tasks <slug> — the roadmap: board tasks planned by the project's boss missions plus chat-planned proposals (status: "proposed"), with result digests, PR links and attempts. Use get_situation <slug> for authoritative verified progress; a completion claim is not verified evidence.

When summarizing, lead with what needs the owner (open decisions, blockers, failed tasks), then progress (summary done/total), then what is running.

Shaping the roadmap from conversation

  • plan_project_tasks — add items. Keys are stable kebab-case, unique per project (task_key). A proposal is a plan, not dispatched work: the project's controller adopts it by planning a real board task under the same key, at which point the proposal drops out automatically.
  • update_project_task — edit an open proposal (title, prompt, acceptance criteria, dependencies). Board tasks already adopted belong to their boss mission — steer the mission instead.
  • cancel_project_task — remove an open proposal.

Prefer small, verifiable items with acceptance criteria over vague epics. Re-planning an existing key updates it in place (idempotent), and revives it if it was cancelled.

Decisions and autonomy

  • answer_project_decision — resolve a pending escalation with the owner's verdict (only when the owner has actually decided in the conversation).
  • record_project_decision — declare an act or escalate a question yourself when you are operating as the project's controller.
  • get_project_grant / set_project_grant — the autonomy grant: level (observe/propose/act_reversible/act_full), merge authority (full | repo:a,b | review-first), budget per tick, parallel missions. Change it only on the owner's explicit request through an operator session. A coordinator may read the grant but cannot broaden it.
Show full SKILL.md (393 more words)Show less

Cross-project etiquette

Unscoped coordinator sessions can read the project roster. Project-scoped sessions can access only their bound project. When the conversation is about one project, do not mutate another one's state without naming it and getting the owner's confirmation first. Reads are always fine.

Workspace Git operations (create_worktree, remove_worktree, merge_branch) require a coordinator/operator session, an explicit mission_id, and an idempotency key. Core routes them to that mission's current machine. Paths stay inside the mission root; use repo_path for a nested checkout. Removal preserves dirty worktrees. Merge requires a clean checkout already on the target branch; inspect conflict/abort evidence before assigning a resolver. Core task-board worktree planning uses the same workspace executor. Task-board scheduling for node/client bosses and desktop-owned workspace operations remain unavailable until their execution routing is migrated. Never substitute a Core path.

An operator may settle an uncertain action with reconcile_action, using the original mission/project scope and concrete evidence from the target. It never replays the action. Do not mark it rejected unless absence of effects is verified.

Workspace jobs follow the mission placement on Core or a node. A remote job receipt includes remote.node_id and spawn_accepted; unknown after a lost response requires inspecting the existing job ID, never a new submission key. Node logs combine stdout/stderr. Cancellation is complete only when the node reports a terminal state. Unsettled jobs prevent moving the workspace.

Reading cloud results economically

get_cloud_execution returns the latest turn by default, omits prompts, and caps each result/detail excerpt at 4096 Unicode characters. Use offset=0 for history. Follow page.next_offset with offset to read new turns; poll the unfinished turn again until it is terminal. For a longer result, use that turn's offset, limit=1, and the field's text_slices.result.next_offset as text_offset. Request include_prompt=true only when needed. Artifact and branch payloads are omitted; counts remain. Do not interpret a page or excerpt boundary as provider completion.

cancel_mission acknowledges a cancellation request, not a stopped process. After its action completes, follow the result's next_tool: use get_cloud_execution for provider confirmation or get_mission_health for native runner termination. Do not report cancellation complete solely because the action is completed or cancel_requested is true.

Gemini CLI (backend="gemini") is retired. Use backend="antigravity" with an exact model from agy models on the execution machine. Existing Gemini CLI missions must be explicitly switched before resuming; do not reuse their native session IDs or treat Google/OpenCode OAuth as an Antigravity login.

© Th0rgal, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/project-manager of Th0rgal/sandboxed.sh.

Open the folder on GitHubat commit f7dcb98

Compare with similar skills

Project Manager next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Project Manager compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Project Manager this skillTh0rgal/sandboxed.sh515—~1.8kAutomated safety check: PassMIT
Jira Natural Language Interfacejjmartres/opencode1333 repos~1.7kAutomated safety check: PassMIT
Fireauto Project Guideimgompanda/fireauto140—~674Automated safety check: PassMIT
Roam Milestone PlanningCranot/roam-code517—~1.2kAutomated safety check: PassApache-2.0
Pm Skillsalirezarezvani/claude-skills28k—~2.6kAutomated safety check: PassMIT
Linearasgeirtj/system_prompts_leaks69k—~653Automated safety check: PassCC0-1.0

Similar skills

  • Lets an agent view, create, update and transition Jira issues in natural language, automatically choosing between the jira CLI and Atlassian MCP tools.

    133 GitHub starsUsed in 3 repos~1.7k tokens
    Product & Project ManagementAuto-check passed
  • Fireauto Project Guide

    imgompanda/fireauto

    프로젝트 생성, 마일스톤 관리, 태스크 추적이 필요할 때 사용하세요. An agent skill from imgompanda/fireauto.

    140 GitHub stars~674 tokensUpdated 6 mo ago
    Product & Project ManagementAuto-check passed
  • Roam Milestone Planning

    Cranot/roam-code

    Define or revise a Roam product milestone and its engineering, adoption and offer-readiness sequence.

    517 GitHub stars~1.2k tokensUpdated 3 days ago
    Product & Project ManagementAuto-check passed
  • Pm Skills

    alirezarezvani/claude-skills

    A skill your agent uses when coordinating project-delivery work across the 8 project-management sub-skills — sprint/velocity analytics, portfolio health, Jira/JQL, Confluence, Atlassian admin…

    28k GitHub stars~2.6k tokensUpdated 1 mo ago
    Product & Project ManagementAuto-check passed
  • Linear

    asgeirtj/system_prompts_leaks

    Read and manage software issues, tickets, projects, and planning data in Linear.

    69k GitHub stars~653 tokensUpdated yesterday
    Product & Project ManagementAuto-check passed
  • Horizon Track

    ruvnet/ruflo

    Track long-horizon objectives across multiple sessions with milestone checkpoints, progress persistence, and drift detection

    74k GitHub stars~744 tokensUpdated today
    Product & Project ManagementAuto-check: notes

More from Th0rgal/sandboxed.sh

All 9 skills in this repo
  • Reads and updates the Sandboxed.sh Library, a git-backed store of skills, agents, commands, tools, rules and MCP servers, through its library tools.

    515 GitHub stars~931 tokensUpdated today
    Auto-check passed
  • Hermes Mission Control

    Th0rgal/sandboxed.sh

    Teaches Hermes to monitor and steer long-running sandboxed.sh missions: spot where a model is stuck, switch backends or models between turns, and send targeted hints.

    515 GitHub stars~8.6k tokensUpdated today
    Auto-check passed
  • Sandboxed.sh Missions

    Th0rgal/sandboxed.sh

    Delegates multi-step coding or research tasks to isolated sandboxed.sh container missions through its MCP tools, each with a chosen workspace, agent profile and prompt.

    515 GitHub stars~23k tokensUpdated today
    Auto-check: notes
  • Orchestrator Executor

    Th0rgal/sandboxed.sh

    Has the agent do the implementation itself and consult one persistent, stronger advisor through ask_worker only when it reaches a dead end.

    515 GitHub stars~620 tokensUpdated today
    Auto-check passed
  • Orchestrator Worker

    Th0rgal/sandboxed.sh

    Sets the rules for a worker agent spawned by a boss mission: stay in scope, verify before finishing, report blockers quickly and end with a clear status.

    515 GitHub stars~531 tokensUpdated today
    Auto-check passed
  • Orchestrator Advisor

    Th0rgal/sandboxed.sh

    Acts as a persistent read-only advisor that maps the repository once, then answers a cheaper executor agent's questions concisely across a long session.

    515 GitHub stars~568 tokensUpdated today
    Auto-check passed

Questions about Project Manager

What does Project Manager do?

Steer sandboxed.sh projects, roadmaps, decisions from chat. An agent skill from Th0rgal/sandboxed.sh. sh.sh projects, roadmaps, decisions from chat.

When should I use Project Manager?

Project Manager fits situations like: tasks that involve Project management.

How do I install Project Manager in Claude Code?

Run `npx skills add Th0rgal/sandboxed.sh --skill project-manager -a claude-code`. Or copy the skill folder (skills/project-manager in Th0rgal/sandboxed.sh) into .claude/skills/project-manager in your project. Claude Code loads it when a task matches its description.

How do I install Project Manager in Codex?

Run `npx skills add Th0rgal/sandboxed.sh --skill project-manager -a codex`. Or copy the skill folder (skills/project-manager in Th0rgal/sandboxed.sh) into .agents/skills/project-manager in your project. Codex loads it when a task matches its description.

Can I use Project Manager in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Th0rgal/sandboxed.sh --skill project-manager -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/project-manager, .gemini/skills/project-manager, .github/skills/project-manager and .opencode/skills/project-manager in your project.

What does Project Manager need to run?

SKILL.md names no scripts, command-line tools or credentials: Project Manager is instructions for the agent only.

Does Project Manager access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Project Manager safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Project Manager use?

Project Manager is published under the MIT licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Project Manager use?

About 1.8k tokens (SKILL.md is roughly 7.2k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Project Manager?

Skills that share tags, products or a category with Project Manager: Jira Natural Language Interface (jjmartres/opencode, 133 stars), Fireauto Project Guide (imgompanda/fireauto, 140 stars), Roam Milestone Planning (Cranot/roam-code, 517 stars) and Pm Skills (alirezarezvani/claude-skills, 28k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Project Manager?

Th0rgal (a GitHub user) maintains it in Th0rgal/sandboxed.sh, which has 515 GitHub stars. The repository holds 9 skills in this directory. The repository was last updated on October 6, 2026.

Source: Th0rgal/sandboxed.sh on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.