Skip to content

Getting Started

Install techne and run your first skill in minutes.

Prerequisites

You'll need:

  • Claude Code: The IDE plugin (visit the marketplace to install if you haven't already)
  • GitHub CLI (gh): Several skills use this to interact with GitHub. Install: github.com/cli/cli
  • A git repository: All skills work within git repos. Initialize one if needed: git init

Installation

From inside Claude Code:

/plugin marketplace add ajbarea/techne
/plugin install techne@techne

Verify the install:

/help techne:audit

This should display the help text for the audit skill. If you see "Command not found", refresh Claude Code and try again.

First Workflow

Let's walk through a typical use case: validating your local build before pushing.

1. Pick Your First Skill

Choose based on what you're trying to do:

Want to... Use
Validate your local build is clean (lint, test, etc.) techne:audit
Organize your pending changes into logical commits techne:auto-commit
Check GitHub Actions for warnings and failures techne:ci-audit
Clean up AI-generated slop in your code techne:deslop
Rewrite docstrings grounded in the code techne:reslop
Verify documentation claims match the code techne:docsync
Maintain your Zensical docs site techne:docs-site
Ground plan decisions in current best practice techne:research-grounded
Audit sister repos for consistency techne:sisters

2. Run Your First Skill

Example: techne:audit to validate the build.

techne:audit

Expected output:

Running techne:audit on <your-repo>

Target: setup ... OK (log matches)
Target: lint  ... OK (log matches)
Target: test  ... OK (log matches)

All targets passed. Your build is clean.

If any target fails or drifts, the skill will highlight it clearly and suggest fixes.

3. Review & Approve

All skills write a plan, report, or diff to disk before making changes. Review it carefully, then approve:

# You see a report or diff
# Review it: does it look right?
# If yes, say "go" or "apply" to execute
# If no, edit manually or ask for refinements

4. Next: Multi-Skill Workflows

Once you're comfortable with one skill, combine them:

# Validate locally
techne:audit

# Organize your changes
techne:auto-commit
# Review COMMITS.md, then say "go"

# Check GitHub Actions output
techne:ci-audit

# Clean up prose before merging
techne:deslop src/

See Examples for real-world workflows.

Configuration

Per-Repo Configuration

Several skills read repo-specific facts from .claude/skill-context.md, one ## section per skill family (markdown headers, not top-level keys). For example:

## audit
phases: [setup, lint, test, ci]
fast_subset: [lint, test]

## slop_ground_truth
authoritative_sources:
  - "src/"

See Conventions for the full skeleton, and Configuration for the user-level ~/.claude/techne.toml.

Multi-Repo Configuration

If you manage multiple repositories (sister repos), create ~/.claude/techne.toml:

github_user = "your-github-username"
workspace_root = "/path/to/your/workspace"

[[sisters]]
name = "repo-one"
status = "active"

[[sisters]]
name = "repo-two"
status = "active"

[[sisters]]
name = "repo-three"
status = "active"

Then use techne:sisters to audit cross-repo consistency. See Configuration for details.

Common Questions

Q: Can I use only one skill?
A: Yes! Each skill is independent. You can use audit without ci-audit, or docsync without any others.

Q: What if a skill tries to modify my code and I disagree?
A: All skills write a plan or diff first. Review it, edit manually if needed, and approve before execution. You're always in control.

Q: Can I customize skill behavior?
A: Yes. Use .claude/skill-context.md for per-repo overrides, and ~/.claude/techne.toml for multi-repo config. See Configuration.

Q: Does techne modify my git history?
A: Only if you approve. Skills like auto-commit stage and commit, but they ask for your permission first. Skills like audit and docsync are read-only by default; they surface issues, you decide whether to fix them.

Q: What's the difference between deslop and reslop?
A: deslop finds and trims AI-slop (verbose, redundant prose). reslop rewrites docstrings grounded in the actual code. Use deslop for triage; use reslop when you want replacement prose, not deletion. See Examples for a real workflow.

Q: How do I debug issues?
A: The skills are prompt-driven, not CLI tools with flags; steer them in natural language. Ask Claude to explain what a skill is doing, or tell it what looked wrong and rerun. If one gets stuck, cancel and try a different approach.

Troubleshooting

"Command not found: techne:audit"
- Ensure you ran /plugin install techne@techne inside Claude Code. - Refresh Claude Code and try again. - Check that you're in a git repository (git status should work).

"gh: command not found"
- Install GitHub CLI: github.com/cli/cli - Authenticate: gh auth login

"Skill runs but output looks wrong"
- Re-run and tell Claude what specifically looked off. There are no --verbose/--debug flags; the skills are prompt-driven, so natural language is the control surface. - Check .claude/skill-context.md for any overrides that might affect behavior. - Ask Claude to explain what the skill is doing.

"Sister repo config not working"
- Verify ~/.claude/techne.toml exists and is readable. - Check the repo names match your actual directories. - Ask techne:sisters which repos it detected from the config.

Next Steps