RED: Research, Evolve, Document
Keep your AI collaborator grounded in what your project has agreed, what is changing, and what still needs investigation.
RED is a methodology for maintaining project knowledge across AI conversations. Use it when an agent confuses an old idea with a current requirement, loses the reasoning behind a design, or leaves you with discussions that never turn into decisions and working changes.
How RED works
| State | What belongs here | How you use it |
|---|---|---|
| Research | Evidence, hypotheses, unknowns, and conflicting information | Investigate questions that could change a decision |
| Evolve | Proposed changes, alternatives, implementation, and acceptance evidence | Develop and review a change before accepting it |
| Document | Accepted goals, behavior, interfaces, and engineering rules | Give people and agents a shared baseline for current work |
Start a task from the relevant Document sources and implementation evidence. Investigate decision-blocking unknowns in Research; a clear change can enter Evolve directly. Review the result before updating Document. Keep small, resolved work in the task; preserve R/E records when work needs handoff, review, or further investigation. Your project chooses whether those records live in Git, local files, issues, or pull requests.
What this package provides
- The RED Skill: instructions that guide a compatible agent through investigation, implementation, review, and documentation. It includes references and artifact templates.
- The
redCLI: project configuration, artifact identifiers, structural checks, Skill installation and updates, and instruction-file management.
The Skill guides the agent's work. The CLI manages files and records decisions you authorize. People or a project-authorized decision process approve transitions; passing a check does not accept a proposal. When code or runtime evidence conflicts with Document, investigate the discrepancy before changing the baseline.
This Python package and the npm distribution provide the same command contract and Skill snapshot. Choose the runtime you already use.
Install the Skill
Requires Python 3.11 or later and pipx. From your project directory:
pipx run --spec red-methodology red skill install --scope repo
This installs the Skill into .agents/skills/red. Use a host that discovers Skills there, and ask the agent to use RED for the project. --scope user installs into your user-level .agents/skills/red instead. Repository scope selects the discovery location; your project decides whether to track or generate the installed files.
The one-shot command runs the installer without adding a permanent red executable to your PATH. For regular CLI use:
pipx install red-methodology
red version
If you already use uv, uvx --from red-methodology red skill install --scope repo is another one-shot option. For a persistent command with uv, use uv tool install red-methodology; upgrade it with uv tool upgrade red-methodology.
Start using RED
After installing the CLI, run these commands from the target project:
red init
# Edit red.toml to map your existing documentation and R/E locations.
red check --json
red status --json
red init creates red.toml and refuses to overwrite an existing configuration. Map the documentation you already maintain; no repository reorganization is required. The Skill can also work without the CLI or configuration when the agent can identify the project's knowledge sources.
Ask your agent, for example:
Use RED to add CSV export. Read the existing requirements, investigate unknowns that affect the design, and show me the implementation and proposed documentation before acceptance.
When an investigation or change needs its own record:
red new research --title "Clarify CSV compatibility requirements"
red new evolve --title "Add CSV export"
These create records in the configured locations. Continue design, implementation, tests, and revision within the agreed scope. At review, the agent presents evidence and the proposed Document update. After acceptance, it synchronizes the maintained documentation. See the CLI contract for promotion commands and JSON output.
Update or remove the Skill
Upgrade the CLI package first, then update the Skill snapshot it carries:
pipx upgrade red-methodology
red skill status --scope repo
red skill update --scope repo
For one-shot use, run pipx run --spec red-methodology red skill update --scope repo. Replacing the runner alone does not update an installed Skill.
red skill uninstall --scope repo
Update and uninstall use the installer's .red-install.json metadata and refuse to alter an unmanaged Skill directory.
Hosts without Skill support
Use a managed block in an existing AGENTS.md, or export standalone instructions:
pipx run --spec red-methodology red instructions install --target AGENTS.md
pipx run --spec red-methodology red instructions export --output RED.md
The managed block preserves the rest of AGENTS.md. Choose the instruction form your host reads; see the adoption guide.
Learn more
- Paper: English · 论文:中文
- RED Protocol 1
- Source and contribution guide
- Report an issue
- Releases, standalone Skill, and plugin archives
Licensed under Apache-2.0.
Metadata
Release files for red-methodology 0.1.2
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| red_methodology-0.1.2.tar.gz | 24.1 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| red_methodology-0.1.2-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 56.2 kB
Release files / red_methodology-0.1.2.tar.gz
| Download URL | red_methodology-0.1.2.tar.gz |
|---|---|
| Size | 24.1 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
3cf0c8891f588cf0c4f35e62c93a3e272a68115a105bf52ec82e88ff37434965
|
|
BLAKE2b-256 checksum How to use checksums |
650ef36e9bb5686e34e2b06c96f4d4e0eaf44d305ae3e00bd177844a884031f2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 11, 2026.
Transparency logRelease files / red_methodology-0.1.2-py3-none-any.whl
| Download URL | red_methodology-0.1.2-py3-none-any.whl |
|---|---|
| Size | 32.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
6a18254d3707365ebb58db68e6c52026fbf2783dc2716422b462a51f7ec3d99d
|
|
BLAKE2b-256 checksum How to use checksums |
aadcc4a59ffd5c1c4ca94fd578a472fcb00f7b2eb6e29f6d44c4faa1f4180aba
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 11, 2026.
Transparency log