Skip to content
Back to skills

Build

ASecurity

Build code in one of three modes — `loop` for fast rounds with the human watching,

  • 3 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 7, 2026
ai-agentspythongoreactgit

Security analysis

A100/100

Pro scans all 9 files and shows the line behind each finding

Scanned November 11, 2026

npx -y skills add buchbend/lore --skill build --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Build?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Build
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/buchbend-build/badge)](https://www.skillsdirectory.com/skills/buchbend-build)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: lore:build
description: Build code in one of three modes — `loop` for fast rounds with the human watching,
  `issue` for one clear issue, `epic` for an epic tracker with a build order. Use when the user
  asks to build, implement, fix, or run an epic, an issue, or quick feedback rounds.
---

# Build

The user names the mode (`/lore:build epic owner/repo#n`), or you pick it from the table.
Follow the mode's file and the rules below. Read another sibling file only when the run reaches it.

| Mode | Use when | Finish point |
|---|---|---|
| [`loop`](loop.md) | The user reacts to a running feature: screenshots, "make this red", "quick fix". | Wrap-up PR from `loop/<slug>` |
| [`issue`](issue.md) | One written, clear issue. | One PR, merged after the user approves |
| [`epic`](epic.md) | An epic tracker with a build-order table from [`to-epic`](../to-epic/SKILL.md). | Epic PR, then the handover section |

Several features, or unshaped work: back to [`orient`](../orient/SKILL.md). One `build` run per epic; a downstream
epic starts after its upstream epic merged.

## Rules for every mode

- **Test first.** Use [`tdd`](../tdd/SKILL.md) for every behaviour a test can observe.
- **Never merge on red.** CI status and ruff come from `gh pr checks <n>`, never from a model. No ruff job in CI: run `ruff check` and `ruff format --check`, read the exit codes. Exception: an `epic` feature branch has no PR and merges on a green local run (`epic.md` § Merge).
- **Decision gate.** An ADR, a PRD or a wiki topic-note edit outside a grilling session is a PR. A human approves its merge. Draft an ADR only after the user approved its ledger line.
- **Merge approval.** A PR that ends a `loop` or `issue` run, or an epic PR, merges only after the user approves it in the conversation. Post one message: the PR link, `gh pr diff <n> --name-only`, the review verdict, `gh pr checks <n>`, the ledger outcomes. Ask whether to merge. On a clear yes, merge with the repo's usual method and report the merge SHA. Silence or a reply about something else is not a yes. The user can merge on GitHub instead; then go on from the merged state.
- **Tiers.** A spawn's model is `lore tier resolve <tier>`; its description opens with its stage tag (`../TIER-DELEGATION.md`). Spawn only the subagents a mode names, or for parallel or isolated work. For one lookup, grep or read directly.
- **Writing rules.** Run `lore style show writing-rules` before you write a PR body, an ADR, a glossary term or the handover section. Write PR bodies and follow-up issues through [`file-issue`](../file-issue/SKILL.md). The board and the verdict block are machine-read and exempt.
- **Ledger.** Log each ADR candidate, term candidate and left-on-the-table item as one `open` line: `lore workflow ledger-add --kind adr|term|left --text "…"`. An ADR candidate meets all three criteria in `../grilling/ADR-FORMAT.md`. A term candidate is a domain word `CONTEXT.md` lacks or uses loosely. `loop`, `issue`: the file is `<git-dir>/lore-ledger.md` of the worktree. `epic`: add `--path -` and paste the printed line into the board's `## Ledger` section.
- **Breakpoint.** After each merged round or feature, add a resume line: `lore workflow ledger-add --kind resume --text "<state>" --next "<next ask>"` (epic: with `--path -`). Then tell the user that `/clear` is safe. After `/clear`, resume from the last resume line.
- **Finish point.** Run `lore workflow ledger-check` (epic: pipe the board comment into `ledger-check -`). Do not finish while it exits 1. Set each outcome with `lore workflow ledger-set <n> --outcome approved|dropped|"filed <owner/repo#n>"` (epic: with `--board -`, `epic-tail.md` § 2).
- **Handover section.** Each finish point writes `## Handover` in the shape [`handover`](../handover/SKILL.md) gives. Work that stops early runs `handover` instead.
- **Logbook.** After the handover section, before the merge approval: `lore log add`, `logbook.md`.
- **Version bump.** If the repo's `AGENTS.md` or `CLAUDE.md` requires a version bump, the shipping PR carries it as its last commit. Lore itself runs `python3 tools/release.py --in-branch --notes notes.md`.
- **Follow-ups.** A defect or a concrete change becomes an issue with a `filed` ledger line. A feature idea outside the run's scope goes to the roadmap: `lore roadmap add` on the build branch. Set its `left` line `approved`. After the merge, run `handover` § Next run.
- **Session end.** Read `lore config get feedback.retrieval_misses`. When true, file one issue per retrieval miss on `feedback.retrieval_misses_repo` through `file-issue`: the fact, the tools tried, the turn count. When false, skip the check. List every issue and PR the session created or commented on in your final message.

Files in this skill

  • SKILL.md5.1 KB
  • epic-tail.md4.5 KB
  • epic.md4.6 KB
  • issue.md1.3 KB
  • logbook.md3.3 KB
  • loop-wrap-up.md3.1 KB
  • loop.md2.9 KB
  • review.md3.2 KB
  • teammate-brief.md2.3 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in ai-agents →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.