Cline Pilot
Delegate Cline CLI coding tasks through a workflow that monitors progress and checks evidence.
- Source repo
- gongdear/cline-pilot
- Stars
- ★ 47
- Last updated
- 13d ago
- License
- MIT
- Primary language
- Python
- FA score
- 41/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificCodex · Claude Code
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Developers using Cline CLI who want to delegate coding tasks to an Agent Skills-compatible assistant while retaining key decisions.
- Not a fit if
- Developers who do not want to use Cline CLI
- Teams expecting the skill to make hard-constraint decisions autonomously
- Source review
- 41/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
Cline Pilot is an Agent Skill that follows the Agent Skills open specification and drives Cline CLI coding tasks on the user's behalf. Its workflow is defined in SKILL.md; it runs Cline CLI and uses scripts/session_report.py to inspect session messages alongside Git and test-report evidence. On first use, it asks about the development environment and branch naming, then stores configuration in the Git-ignored references/local-config.md. It relays decision points to the user in a fixed four-part format and checks a six-item acceptance checklist before reporting completion. The skill does not hold project architecture knowledge and does not independently push code, delete files, write to databases, spend money, or change global configuration.
Once installed, a compatible Agent loads SKILL.md when activated and selects an interactive or non-interactive Cline CLI run mode for the task. On first use, it asks for settings such as the conda/Python environment, how the toolchain is added to PATH, and task branch naming, then writes them to private references/local-config.md. Cline runs inside the user's development environment; for background tasks, scripts/session_report.py reads session files and monitors progress using Git status and test reports. The skill relays pending user decisions in a fixed four-part format and records corrections and decisions by project tag class. After at least two consistent samples, it distills stable preferences into SKILL.md; before reporting completion, it checks a six-item acceptance checklist.
- Developers using Cline CLI who want to delegate coding tasks to an Agent Skills-compatible assistant while retaining key decisions.
- Teams that need Cline tasks to run in their own conda or Python environment with environment and branch conventions recorded.
- Engineers monitoring long-running background coding tasks who want to check session files, Git status, and test reports.
- Teams starting a new project that want to gather rules from the user one dimension at a time.
- Teams with an existing codebase that want project rules grounded in a code scan.
- Individual developers who want to accumulate preferences by project type and reduce repeated decisions.
How do you install or deploy this agent?
Install with the npx skills CLI, or copy the skill directory into a compatible Agent's skills directory. First use requires details such as the development environment, how the toolchain reaches PATH, and task branch naming; see references/local-config.example.md for a configuration template. The README does not document how to install Cline CLI, its version requirements, or model configuration.
npx skills add https://github.com/gongdear/cline-pilotAlternatively, copy the skill directory, for example into Hermes:
mkdir -p ~/.hermes/skills && cp -r cline-pilot ~/.hermes/skills/How do you use this agent?
Enable the skill in an Agent that supports Agent Skills. On first use, answer the questions about your development environment and branch naming; then give the Agent a coding task for Cline CLI. While it runs, use the script to inspect session and Git/test evidence:
python3 scripts/session_report.py 15 /path/to/repoReplace 15 and /path/to/repo as needed. The README does not give the full parameter definitions or a specific Cline CLI launch command.
What are this agent's strengths and limitations?
- Checks progress against session files, Git status, and test reports instead of relying only on an Agent's own claims.
- Relays decision points in a fixed four-part format and leaves hard-constraint actions for the user to decide.
- Keeps user-specific configuration and preferences in Git-ignored private files; public files and templates carry no user-specific secrets.
- Provides separate rule-initialization paths for new projects and existing codebases.
- The monitoring script uses Python's standard library and can run independently.
- Task execution depends on Cline CLI; the README does not document an alternative execution engine.
- First-time setup requires environment, PATH, and branch-naming details.
- The README does not specify Cline CLI installation, version, model, or cost requirements.
- Preference distillation requires at least two consistent samples, so early tasks still need user decisions.
- The available material does not enumerate the six acceptance-checklist items.
How does this agent compare with similar options?
The README presents it as an Agent Skill that invokes Cline CLI and says it works with supporting hosts such as Hermes, Cline, Claude Code, Codex, Cursor, and OpenCode. These are host options, not a feature comparison with alternative products.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Cline Pilot This agent | 41 · Major gaps | Agent plugin / skill | ★ 47 | 13d ago | Python | Codex · Claude Code |
| GameDesignOS: A Local-First Operating System for AI-Assisted Game Design | 78 · Good | CLIFree | ★ 426 | 2mo ago | Python | Codex · Claude Code |
| EOC — Everything OpenAI Codex Workflow OS | 75 · Good | Agent plugin / skillFree + model costs | ★ 86 | 4mo ago | JavaScript | Codex |
| Keep the Why | 75 · Good | Agent plugin / skillFree | ★ 231 | 1d ago | Python | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The README says the skill asks the user about hard-constraint decisions and always asks before pushing, deleting, writing to a database, spending money, or changing global configuration. This is partial evidence for confirmation and external-effect controls, but no workflow files are provided to corroborate it, so full marks are not justified. The README also says private configuration is git-ignored and public files contain no user-specific secrets, but gives little detail on actual data handling or access controls; the read-only monitor script and dependency security cannot be checked from the supplied files. No rollback mechanism is described. Attribution is limited to a copyright name in the license, and publisher identity remains unverified.
The README describes monitoring through session files and git/test-report evidence, and calls the script deterministic and standard-library-only; these are useful claims but are not corroborated by code or test material. Dependency availability is covered only by mentioning the standard library and several external CLIs/tools, with no version or availability guarantees. The supplied material does not describe failure messages or recovery behavior.
The README identifies support for agents implementing the Agent Skills specification and describes Cline CLI coding tasks, with separate cold-start and existing-code scenarios. It also says the user is asked about the conda/Python environment, PATH, and branch naming, which gives some evidence of environment fit. Trigger conditions are summarized only as coding tasks targeting Cline CLI, with no specific boundaries or handling for accidental activation. Capability boundaries are listed, but the core skill file that would substantiate them was not supplied.
The README provides a directory layout, progressive-disclosure guidance, two installation methods, a quick start, and validation commands. The examples are workflow snippets rather than a FAQ. Some naming and contribution conventions are stated, but there is no version or changelog and no clear maintainer or update commitment. The LICENSE file supplies the full MIT terms, so license evidence is strong.
The README describes a fixed format for relaying decision points, a six-item acceptance checklist, and a session-monitoring command, suggesting outputs intended to support user action. However, utility and comparative value are asserted without implementation or outcome evidence. The model advice mentions a paid strong model and a local model, but provides no cost figures or comparison evidence, so cost-benefit earns only partial credit.
The README points to git status, test-report numbers, and non-empty artifacts as completion evidence, and provides a script invocation and validation commands. These offer some traceability, but this review has only the README and license, with no code, tests, or other sources for cross-checking. The document partly distinguishes design claims from suggested validation, but does not label which statements are verified facts, maintainer recommendations, or unverified claims.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- The supplied material contains only the README and MIT license. Claims about confirmation flows, privacy separation, read-only monitoring, determinism, and acceptance evidence are not verified against the skill files or script.
- Before use, check the actual skill workflow and local configuration ignore rules, and confirm that access to external CLIs and test reports matches expectations.
FAQ
Does it require a paid model or API key?
Will it push code or carry out hard-constraint actions on my behalf?
How can I tell whether a background task is complete?
python3 scripts/session_report.py 15 /path/to/repo.Can it remember preferences across projects?
SKILL.md after at least two consistent samples.