Files

kaykayyali d105928204 init: lore skill — agent interface to lore.wiki.git

Adds the 'lore' skill: an interface for any agent to read and write
the shared homelab wiki at git.homelab.local/kaykayyali/lore.wiki.git.

The skill teaches agents:
  * when to write to the wiki vs local memory (rule of thumb: if
    another agent on a different project would benefit, wiki)
  * the wiki's own protocol (Schema + Write Protocol + Agent
    Onboarding) — read those first, follow them exactly
  * how to draft, commit, and push via local clone + SSH
  * that git add -A is forbidden (dev cruft has burned the wiki)

Helper scripts:
  * lore-sync.sh — clone or pull --rebase the wiki
  * lore-add.sh  — stage specific files, commit, push (prompts first)

The skill follows the wiki's tag taxonomy and append-only model:
local memory is for the current session; the wiki is for everything
worth remembering.

Co-Authored-By: Claude <noreply@anthropic.com>

2026-06-23 15:02:27 -04:00

2.8 KiB

Raw Permalink Blame History

lore

The shared knowledge base for the homelab. Two parts:

lore.wiki.git — the wiki itself. The cross-project, append-only source of truth for decisions, pitfalls, patterns, entities, and project state. Catalog: https://git.homelab.local/kaykayyali/lore/wiki
lore.git (this repo) — the agent skill that interfaces with the wiki. Install once per agent and it knows how to read, draft, commit, and push wiki pages following the wiki's own protocol.

What goes in the wiki

Anything an agent on a different project might need in the future. Decisions with rationale. Pitfalls (something broke, here's why). Patterns (something worked, capture it). New projects, services, facets. Project state.

What stays in local memory

Session-private notes. Working memory for the current task. Drafts of pages before they're approved. Anything project-private.

Rule of thumb: if another agent on a different project would benefit, it belongs in the wiki. Local memory is for the current session only.

Install the skill

Copy or symlink the skills/lore/ directory into your agent's skills directory.

For Claude Code:

# From this repo
ln -s "$(pwd)/skills/lore" ~/.claude/skills/lore

For other agents, follow their skill-loading convention. The skill file is skills/lore/SKILL.md — the name: lore frontmatter and the description: ... block tell the agent when to load it.

Use

# Pull the latest wiki into ~/lore-wiki (clones on first run)
./skills/lore/scripts/lore-sync.sh

# Read first
less ~/lore-wiki/Schema.md
less ~/lore-wiki/Write\ Protocol.md
less ~/lore-wiki/Agent\ Onboarding.md

# Write — follow the protocol exactly
# Draft files in ~/lore-wiki/, update Index.md and Log.md,
# show the user the diff, then:
./skills/lore/scripts/lore-add.sh "<file>" [<file>...]

The skill's SKILL.md has the full workflow and rule set.

Repo layout

lore.git/
├── README.md                       # this file
└── skills/
    └── lore/
        ├── SKILL.md                # the agent skill
        ├── references/
        │   └── QUICKREF.md         # condensed Schema for fast drafting
        └── scripts/
            ├── lore-sync.sh        # clone / pull --rebase
            └── lore-add.sh         # stage + commit + push

2.8 KiB Raw Permalink Blame History