capiko-ai
llms.txt2 starsChanged 20 days agoMITView on GitHub
capiko-ai turns the GitHub Copilot CLI into a senior teammate that already knows your team's playbook. It installs a curated catalog of skills, a teaching-first persona, cross-session memory (engram), and a full Spec-Driven Development (SDD) workflow into ~/.copilot. capiko directs; Copilot executes — capiko never calls a model itself, so everything it installs is reproducible and auditable. It targets GitHub Copilot only and ships as a single native Go binary (a Bubbletea TUI plus headless subcommands). Configs land in ~/.copilot/;…
- Pipes a download into a shell
- Deletes or force-pushes
- Installs packages
- Commits and pushes
We split this llms.txt into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 24
Run with no arguments to launch the interactive TUI.
≈ “Common” is an estimate: at least 1 in 5 of the 3,225 llms.txt files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
# capiko-ai > capiko-ai turns the GitHub Copilot CLI into a senior teammate that already knows > your team's playbook. It installs a curated catalog of skills, a teaching-first > persona, cross-session memory (engram), and a full Spec-Driven Development (SDD) > workflow into `~/.copilot`. capiko **directs**; Copilot **executes** — capiko > never calls a model itself, so everything it installs is reproducible and auditable. It targets GitHub Copilot only and ships as a single native Go binary (a Bubbletea TUI plus headless subcommands). Configs land in `~/.copilot/`; state, backups, and drift tracking live in `~/.capiko/`. Module path: `github.com/martinhg/capiko-ai`. License: MIT. ## Install - Homebrew (macOS/Linux): `brew install martinhg/homebrew-tap/capiko-ai` - Script (macOS/Linux): `curl -sL https://raw.githubusercontent.com/martinhg/capiko-ai/main/scripts/install.sh | bash` - PowerShell (Windows): `irm https://raw.githubusercontent.com/martinhg/capiko-ai/main/scripts/install.ps1 | iex` - Scoop (Windows): `scoop bucket add capiko-ai https://github.com/martinhg/scoop-bucket && scoop install capiko-ai` - From source (Go 1.26+): `go install github.com/martinhg/capiko-ai/cmd/capiko-ai@latest` ## CLI surface Run with no arguments to launch the interactive TUI. Headless subcommands: - `capiko-ai install [--all] [--json]` — install skills, agents, persona, and configs non-interactively - `capiko-ai sync [--json]` — reconcile installed configs with the catalog (auto-repair drift); warns when a managed engram binary is outdated - `capiko-ai uninstall [--json]` — remove capiko-managed configs - `capiko-ai doctor [--json] [--repair]` — ecosystem diagnostics (exit non-zero on failure); --repair re-applies the managed catalog when drift is found - `capiko-ai sdd-status [--json]` — deterministic SDD workflow state for a change - `capiko-ai sdd-continue` — advance the next dependency-ready SDD phase - `capiko-ai skill-registry` — index installed skills by trigger and path - `capiko-ai version` — print the version Exit codes for headless commands: 0 success, 1 error, 2 Copilot not found. ## What capiko installs - **Skills** (`~/.copilot/skills/<name>/SKILL.md`) — 21 embedded skills Copilot auto-discovers: the 10-skill SDD pipeline (explore, propose, spec, design, tasks, apply, verify, archive, init, onboard) plus `sdd-shared`, `skill-creator`, `skill-registry`, `codebase-docs`, the 4R review set (`review-risk`, `review-readability`, `review-reliability`, `review-resilience`), and the capiko house skills (`capiko-conventions`, `capiko-pr`, `capiko-hello`). - **Copilot agents** (`~/.copilot/agents/<name>.agent.md`) — 8 SDD-phase agents plus a coordinator; the real delegation targets the SDD workflow routes to. - **Persona** (`~/.copilot/copilot-instructions.md`) — a teaching-first (Rioplatense), Neutral, or None instruction block shaping how Copilot talks and reasons. - **Scoped instructions** (`~/.copilot/instructions/*.instructions.md`) — applied per matching file via `applyTo` globs. - **Engram memory** (`~/.copilot/mcp-config.json`, VS Code `mcp.json`) — wires the [engram](https://github.com/Gentleman-Programming/engram) MCP server for cross-session memory (local or `hybrid` cloud). - **Team memory sync** (`.git/hooks/post-merge` + `.git/hooks/pre-push`, opt-in) — writes marker-delimited hook blocks so `post-merge` runs `engram sync --import` and `pre-push` runs `engram sync --project '<name>'`, sharing Engram memory across the team through git with no cloud required. Requires explicit ack of the scope-leak risk (all project memories including `scope:personal` are committed to git). Mitigations: wrap sensitive content in `<private>…</private>` tags, or use a separate project name for personal memories. If a competing hook framework (`core.hooksPath`, husky, lefthook, pre-commit) is detected, writes are skipped and the equivalent shell lines are shown for manual integration. capiko never installs the `engram` binary. - **Copilot hooks (guardrails)** (`$COPILOT_HOME/hooks/capiko-guardrails.json`, opt-in) — a capiko-owned `preToolUse` hook that inspects each tool call and, per posture, asks (`warn`) or denies (`strict`) before four dangerous patterns: `rm -rf /`, pipe-to-shell (`curl | sh`), `git push --force` to main/master, and `chmod 777`. Non-matches are allowed silently. User-level only (repo-level `.github/hooks` is future); org policy hooks can override; a `timeoutSec` overrun fails open. v1 interpolates no user input, so the shell-injection surface is empty. Re-applied on every sync. - **Safety net** (`~/.capiko/`) — snapshot-before-mutate backups, per-skill checksum state, drift detection, and self-update. ## Architecture notes - capiko is an **install-time file writer**, not a runtime proxy or daemon. It writes static configs Copilot auto-discovers; it never intercepts the request path. - It ships its own **native SDD engine** in Go (`internal/sddstatus`): `sdd-status` and `sdd-continue` compute workflow state deterministically rather than asking a model. - It **configures** engram (never provisions/downloads it), the same boundary it keeps for any external tool. ## Docs - [README](https://github.com/martinhg/capiko-ai/blob/main/README.md) — full overview, quickstart, menu, SDD workflow - [Team memory sync](https://github.com/martinhg/capiko-ai/blob/main/README.md#-team-memory-sync) — git hook sharing, scope-leak warning, conflict coexistence - [Copilot hooks (guardrails)](https://github.com/martinhg/capiko-ai/blob/main/README.md#-copilot-hooks-guardrails) — posture levels, the four patterns, scope and fail-open caveats - [Headless CLI](https://github.com/martinhg/capiko-ai/blob/main/README.md#-headless-cli) — scripting, CI, dotfiles - [Spec-Driven Development](https://github.com/martinhg/capiko-ai/blob/main/README.md#-spec-driven-development) — the 8-phase pipeline - [How it works](https://github.com/martinhg/capiko-ai/blob/main/README.md#-how-it-works) — the direct/execute split - [Releases](https://github.com/martinhg/capiko-ai/releases) — changelog and pre-built binaries
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- branch-prSkill
- capiko-devSkill
- chained-prSkill
- cognitive-doc-designSkill
- comment-writerSkill
- go-testingSkill
- issue-creationSkill
- work-unit-commitsSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

