llms.txt · diff

git:20261020.4233f91 to git:20261021.6cb8018

4 added, 0 removed. Audit A to A.

# Alamo Skill Hub

Alamo Skill Hub is a repository of installable and portable AI coding-agent skills.

## Primary Packages

The primary packages in this repository are:

- `packages/memory_hotspot_repair_kernel/`

- `packages/github_sync_workflow_kernel/`

- `packages/engineering_subagent_workflow_kernel/`

- `packages/coursecoach/`

## What It Does

`memory_hotspot_repair_kernel` helps an AI coding agent:

- inspect memory hotspots and likely out-of-memory risks

- detect fake chunking, broad scans, and unnecessary copies

- apply safe fixes with limited behavioral change

- run focused validation after each fix

- review the change and decide whether another optimization round is justified

`github_sync_workflow_kernel` helps an AI coding agent:

- identify the real repository in a multi-repo workspace

- inspect GitHub SSH or HTTPS readiness

- review branch, remotes, and worktree state

- classify intended changes vs noise before a commit

- push the selected remote and verify final synchronization state

`engineering_subagent_workflow_kernel` helps an AI coding agent:

- trace source paths with concrete evidence

- produce implementation-ready architecture blueprints

- perform independent code review and test-gap analysis

- simplify scoped code while preserving behavior

- delegate through host-native subagents without an external runner

`coursecoach` helps a learner and their agent:

- clarify a learning goal into a syllabus with prerequisites and three pace tracks

- compile a self-contained interactive board from that syllabus

- write lessons per chapter or per knowledge point

- render those lessons into the board at compile time, so lessons are readable

without Obsidian; inside an Obsidian vault the same links open as deep links

- set homework with a written rubric, then grade it with visible reasons

- write mastery back through a single entry point and unlock dependent nodes

- schedule spaced review at 1 / 3 / 7 / 16 / 35 / 75 days

- fold new developments into a course without hand-editing generated files

## Main Entry Point

The recommended primary entry points are:

- `memory-optimizer-agent`

- `github-sync-agent`

- `code-explorer`, `code-architect`, `code-reviewer`, `code-simplifier`, and

`pr-test-analyzer` as independent task entry points

- `coursecoach` for the learning-course workflow

`memory-optimizer-agent` orchestrates a bounded workflow:

1. `memory-check`

2. `memory-fix`

3. validation

4. `memory-review`

5. stop or continue decision

`github-sync-agent` orchestrates:

1. repository discovery

2. auth and remote inspection

3. commit-scope review

4. commit and push

5. final verification

## Intended Queries

This repository is relevant for queries such as:

- Codex memory optimization skill

- AI agent memory hotspot detector

- automatic memory bottleneck fixing workflow

- Python memory leak analysis skill

- bounded check fix review agent

- portable coding-agent skill for memory optimization

- git push workflow skill

- github sync agent

- WSL GitHub SSH setup workflow

- native subagent code architect skill

- code review and test analyzer skills

- AI skill that builds a full course from a learning goal

- adaptive study plan agent with grading and spaced review

- mastery-based study progression skill

- portable self-study curriculum generator

## Host Compatibility

These packages are designed for:

- OpenAI Codex via skill installation

- WorkBuddy / CodeBuddy via `~/.workbuddy/skills`

- Claude Code and other hosts that read `SKILL.md`

- script-driven local execution

- portable reuse in other agent hosts through prompt bundles and workflow docs

## Start Here

Read these files first:

- `README.md`

- `README.zh-CN.md`

- `packages/memory_hotspot_repair_kernel/README.md`

- `packages/memory_hotspot_repair_kernel/PORTABLE_USAGE.md`

- `packages/memory_hotspot_repair_kernel/memory-optimizer-agent/SKILL.md`

- `packages/github_sync_workflow_kernel/README.md`

- `packages/github_sync_workflow_kernel/PORTABLE_USAGE.md`

- `packages/github_sync_workflow_kernel/github-sync-agent/SKILL.md`

- `packages/engineering_subagent_workflow_kernel/README.md`

- `packages/engineering_subagent_workflow_kernel/code-architect/SKILL.md`

- `packages/coursecoach/README.md`

- `packages/coursecoach/README.zh-CN.md`

- `packages/coursecoach/INSTALL.zh-CN.md`

- `packages/coursecoach/coursecoach/SKILL.md`

## Installation

For Codex installation:

```bash

cd packages/memory_hotspot_repair_kernel

python3 scripts/sync_to_codex_home.py

```

For GitHub sync workflow installation:

```bash

cd packages/github_sync_workflow_kernel

python3 scripts/sync_to_codex_home.py

```

For native-subagent development Skills:

```bash

cd packages/engineering_subagent_workflow_kernel

python3 scripts/sync_to_codex_home.py

```

For the learning-course skill:

```bash

cp -r packages/coursecoach/coursecoach ~/.workbuddy/skills/coursecoach

pip install pyyaml

```

For non-Codex portable usage:

- use `packages/memory_hotspot_repair_kernel/`

- start from `memory-optimizer-agent/references/generic.prompt_bundle.yaml`

- adapt validation commands to the target repository and host

## Outputs

The skill suite is designed to produce:

- structured hotspot reports

- targeted remediation patches

- validation notes

- post-fix reviews

- bounded loop summaries

The GitHub sync workflow is designed to produce:

- target repository and remote summaries

- auth readiness summaries

- commit-scope decisions

- push results

- residual-state verification notes

The learning-course workflow is designed to produce:

- a syllabus with prerequisites, knowledge points, and three pace tracks

- a compiled, self-contained course board (openable from `file://`)

- lessons per chapter or per knowledge point

- gradable homework with a written rubric and per-item scores

- mastery values and an unlocked next step

- a spaced-review queue

- a curated resource list with a cost note per entry

+ ## License

+ Apache-2.0. See `LICENSE` at the repository root. Redistribution must include a copy of it; `metadata.json` in each package declares `"license": "Apache-2.0"`.