build

Name: build
Author: JuliusBrussee

2.4kinstalls81forks

How to install build?

juliusbrussee/cavekit/build

$npx -y skills add JuliusBrussee/cavekit --skill build

Installs into the current project.

Summary

Plan-then-execute implementation against SPEC.md. Native single-thread loop, no sub-agents. On test or build failure, auto-invokes the backprop skill before retrying — a failed verification always considers whether a new §V invariant would prevent recurrence. Triggers when the us

Files · 1

View on GitHub

SKILL.md

1	# build — implement spec
2
3	Single-thread native plan→execute. You are main Claude. No swarm.
4
5	## LOAD
6
7	1. Read `SPEC.md`. If missing → tell user to invoke the spec skill first. Stop.
8	2. Read `FORMAT.md` once if not loaded.
9	3. Read §R if present — external facts the build must honor, ⊥ re-derive or contradict.
10	4. Parse invocation args:
11	- `§T.n` → that task only
12	- `--next` → lowest-numbered row with status `.` or `~`
13	- `--all` or empty → every `.` row in §T order
14
15	High blast radius (shared module, auth, data, money, public §I)? Run `/review` first. Trivial & reversible? Skip planning ceremony, just do step EXECUTE.
16
17	## PLAN
18
19	Native plan mode — you delegate to it, you do not reinvent task breakdown. For chosen task(s):
20
21	1. Cite every §V invariant that applies. Plan must respect all.
22	2. Cite every §I interface touched. Plan must preserve shape.
23	3. List files to create / edit.
24	4. Verification contract — name the EXACT test(s) / acceptance criteria that
25	prove each §V touched. Which test, not "add tests". "Do TDD" alone backfires;
26	the spec says what to check. Each §V touched → a named test that fails first.
27	5. Name verification command (test, build, lint) — this is the external oracle. Green = done; ⊥ "looks done".
28
29	Show plan. Wait for user OK unless auto mode.
30
31	## EXECUTE
32
33	Per task in order:
34
35	1. Flip §T.n status cell `.` → `~`. Just write to SPEC.md.
36	2. Edit code per plan.
37	3. Run verification command.
38	4. Pass → flip `~` → `x`. Next task.
39	5. Fail → invoke backprop skill. Do NOT retry blindly.
40
41	## FAIL → BACKPROP
42
43	On test/build failure:
44
45	1. Read failure output.
46	2. Ask: is failure (a) my code bug, (b) spec wrong, or (c) unspecified edge case?
47	3. If (a) → fix code, re-run. No spec change.
48	4. If (b) or (c) → invoke spec skill with `bug: <cause>` first, let it update §V and §B, then resume build against updated spec.
49
50	Rule: never silently fix root-cause without considering backprop. §B is the memory that stops recurrence.
51
52	## WRITE POLICY
53
54	- Only flip §T status. No other SPEC.md edits from build.
55	- Other spec edits → invoke spec skill.
56	- Commit after each §T completes. Message: `T<n>: <goal line>` + §V cites.
57
58	## VERIFICATION
59
60	Task `x` only if:
61	- Verification command (the oracle) exits 0.
62	- Every §V touched has its named test from the verification contract, and it passes.
63	- No §V invariant regressed (run full test suite at end).
64
65	## NON-GOALS
66
67	- No sub-agents. No parallel workers. Main thread only.
68	- No progress dashboards. `cat SPEC.md \| grep §T` is the dashboard.
69	- No speculative work beyond chosen task scope.

People who liked this also viewed

6 picks

Type

build

2.4kinstalls81forks

View on GitHub

How to install build?

juliusbrussee/cavekit/build

$npx -y skills add JuliusBrussee/cavekit --skill build

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# build — implement spec
2
3	Single-thread native plan→execute. You are main Claude. No swarm.
4
5	## LOAD
6
7	1. Read `SPEC.md`. If missing → tell user to invoke the spec skill first. Stop.
8	2. Read `FORMAT.md` once if not loaded.
9	3. Read §R if present — external facts the build must honor, ⊥ re-derive or contradict.
10	4. Parse invocation args:
11	- `§T.n` → that task only
12	- `--next` → lowest-numbered row with status `.` or `~`
13	- `--all` or empty → every `.` row in §T order
14
15	High blast radius (shared module, auth, data, money, public §I)? Run `/review` first. Trivial & reversible? Skip planning ceremony, just do step EXECUTE.
16
17	## PLAN
18
19	Native plan mode — you delegate to it, you do not reinvent task breakdown. For chosen task(s):
20
21	1. Cite every §V invariant that applies. Plan must respect all.
22	2. Cite every §I interface touched. Plan must preserve shape.
23	3. List files to create / edit.
24	4. Verification contract — name the EXACT test(s) / acceptance criteria that
25	prove each §V touched. Which test, not "add tests". "Do TDD" alone backfires;
26	the spec says what to check. Each §V touched → a named test that fails first.
27	5. Name verification command (test, build, lint) — this is the external oracle. Green = done; ⊥ "looks done".
28
29	Show plan. Wait for user OK unless auto mode.
30
31	## EXECUTE
32
33	Per task in order:
34
35	1. Flip §T.n status cell `.` → `~`. Just write to SPEC.md.
36	2. Edit code per plan.
37	3. Run verification command.
38	4. Pass → flip `~` → `x`. Next task.
39	5. Fail → invoke backprop skill. Do NOT retry blindly.
40
41	## FAIL → BACKPROP
42
43	On test/build failure:
44
45	1. Read failure output.
46	2. Ask: is failure (a) my code bug, (b) spec wrong, or (c) unspecified edge case?
47	3. If (a) → fix code, re-run. No spec change.
48	4. If (b) or (c) → invoke spec skill with `bug: <cause>` first, let it update §V and §B, then resume build against updated spec.
49
50	Rule: never silently fix root-cause without considering backprop. §B is the memory that stops recurrence.
51
52	## WRITE POLICY
53
54	- Only flip §T status. No other SPEC.md edits from build.
55	- Other spec edits → invoke spec skill.
56	- Commit after each §T completes. Message: `T<n>: <goal line>` + §V cites.
57
58	## VERIFICATION
59
60	Task `x` only if:
61	- Verification command (the oracle) exits 0.
62	- Every §V touched has its named test from the verification contract, and it passes.
63	- No §V invariant regressed (run full test suite at end).
64
65	## NON-GOALS
66
67	- No sub-agents. No parallel workers. Main thread only.
68	- No progress dashboards. `cat SPEC.md \| grep §T` is the dashboard.
69	- No speculative work beyond chosen task scope.

People who liked this also viewed

6 picks

Type