run

Name: run
Author: Borda

23stars3forks

How to install run?

borda/ai-rig/run

$npx -y skills add Borda/AI-Rig --skill run

Installs into the current project.

Summary

Sustained metric-improvement loop with atomic commits, auto-rollback, and experiment logging. Iterates with specialist agents, commits atomically, auto-rolls back on regression. Accepts a program.md file path. Supports --resume, --team, --colab, --codex, --researcher, --architect

Files · 1

View on GitHub

SKILL.md

1	<objective>
2
3	Sustained metric-improvement loop — reads `program.md`, iterates specialist ideation agents, commits atomically, auto-rolls back on regression. For long-running automated improvement campaigns.
4
5	NOT for: methodology validation before run (use `/research:judge`); hypothesis generation (use `research:scientist` agent); one-off feature work (use `/develop:feature`).
6
7	</objective>
8
9	<constants>
10
11	Campaign mode only:
12
13	```yaml
14	MAX_ITERATIONS: 50 (hard cap); DEFAULT 20 when max_iterations unset in program.md; program.md may raise up to 50; values above 50 clamped to 50 with a warning
15	MAX_CODEX_RUNS: 10 (cost ceiling for --codex Phase 2c — disable Codex once exceeded)
16	STUCK_THRESHOLD: 5 consecutive discards → escalation
17	GUARD_REWORK_MAX: 2 attempts before revert
18	VERIFY_TIMEOUT_SEC: 120 (local), 300 (--colab)
19	COLAB_KNOWN_HW: H100, L4, T4, A100
20	SUMMARY_INTERVAL: 10 iterations
21	DIMINISHING_RETURNS_WINDOW: 5 iterations < 0.5% each → warn user and suggest stopping
22	STATE_DIR: .experiments/state/<run-id>/ (timestamped dir per run — see .claude/rules/artifact-lifecycle.md)
23	SENTINEL_SLUG_FORMULA: \|
24	eval "$(bash "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/git_slugs.sh")"
25	# Sentinel path: ${TMPDIR:-/tmp}/claude-commit-auth-${REPO_SLUG}-${BRANCH_SLUG} # tmpdir-exempt: user-shell-boundary
26	# Bash state is lost between tool calls — re-source git_slugs.sh at each use site; it is the only authorized slug form.
27	```
28
29	<!-- Note: STATE_DIR (.experiments/state/) holds per-iteration artifacts (diary, experiments.jsonl).
30	Hypothesis pipeline outputs (hypotheses.jsonl, checkpoint.json, journal.md) go to .experiments/<run-id>/ (RUN_DIR).
31	These are two separate directories by design — see protocol.md for layout. -->
32
33	Agent strategy mapping (`agent_strategy` in config → ideation agent to spawn):
34
35	\| `agent_strategy` \| Specialist agent \| When to use \|
36	\| --- \| --- \| --- \|
37	\| `auto` \| heuristic \| Default — infer from metric_cmd keywords \|
38	\| `perf` \| `foundry:perf-optimizer` \| latency, throughput, memory, GPU utilization \|
39	\| `code` \| `foundry:sw-engineer` \| coverage, complexity, lines, coupling \|
40	\| `ml` \| `research:scientist` \| accuracy, loss, F1, AUC, BLEU \|
41	\| `arch` \| `foundry:solution-architect` \| coupling, cohesion, modularity metrics \|
42
43	Auto-inference keyword heuristics (when `agent_strategy: auto` or omitted; checked against `## Goal` text AND metric command):
44
45	Precedence order (first match wins; ML keywords beat test-framework keywords). ML-specific compound terms (not bare tokens) required — prevents over-triggering on `eval`/`train`/`val` as common words:
46	- contains `accuracy`, `loss` (paired with `train_loss`/`val_loss`/`eval_loss`), `f1_score`, `auc_roc`, `auroc`, `train_step`, `val_acc`, `eval_loss`, `epoch`, `gradient`, `tensor`, `overfit`, `generaliz`, `regulariz`, `validation`, `dropout`, `weight_decay`, `lr_schedule`, `cross_val`, `precision`, `recall`, OR explicit `--scientist` flag → `ml` → `research:scientist`
47	- contains `time`, `latency`, `bench`, `throughput`, `memory` → `perf` → `foundry:perf-optimizer`
48	- contains `pytest`, `coverage`, `complexity` → `code` → `foundry:sw-engineer`
49	- no keyword match → `perf` (default fallback) — WARN: print `⚠ No keyword match — defaulting to 'perf' strategy. If this is an ML task, set agent_strategy: ml in program.md.` Log resolved agent + reason in state.json `strategy_resolution`.
50
51	Bare tokens `eval`, `train`, `val` (without compound suffix) do NOT trigger `ml` routing — too common in non-ML contexts (test eval scripts, training-environment configs, validator command names).
52
53	Stuck escalation sequence (at STUCK_THRESHOLD consecutive discards):
54
55	1. Switch agent type. Rotation by current strategy:
56
57	\| Current strategy \| Next strategy \| Escalation agent \|
58	\| --- \| --- \| --- \|
59	\| `code` \| `ml` \| `research:scientist` \|
60	\| `ml` \| `perf` \| `foundry:perf-optimizer` \|
61	\| `perf` \| `code` \| `foundry:sw-engineer` \|
62	\| `arch` \| `code` \| `foundry:sw-engineer` (fallback `foundry:solution-architect` if sw-engineer unavailable) \|
63	\| `auto` \| infer from resolved strategy \| follow rotation row for whichever concrete strategy `auto` heuristics resolved t

People who liked this also viewed

6 picks

Type

run

23stars3forks

View on GitHub

How to install run?

borda/ai-rig/run

$npx -y skills add Borda/AI-Rig --skill run

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	<objective>
2
3	Sustained metric-improvement loop — reads `program.md`, iterates specialist ideation agents, commits atomically, auto-rolls back on regression. For long-running automated improvement campaigns.
4
5	NOT for: methodology validation before run (use `/research:judge`); hypothesis generation (use `research:scientist` agent); one-off feature work (use `/develop:feature`).
6
7	</objective>
8
9	<constants>
10
11	Campaign mode only:
12
13	```yaml
14	MAX_ITERATIONS: 50 (hard cap); DEFAULT 20 when max_iterations unset in program.md; program.md may raise up to 50; values above 50 clamped to 50 with a warning
15	MAX_CODEX_RUNS: 10 (cost ceiling for --codex Phase 2c — disable Codex once exceeded)
16	STUCK_THRESHOLD: 5 consecutive discards → escalation
17	GUARD_REWORK_MAX: 2 attempts before revert
18	VERIFY_TIMEOUT_SEC: 120 (local), 300 (--colab)
19	COLAB_KNOWN_HW: H100, L4, T4, A100
20	SUMMARY_INTERVAL: 10 iterations
21	DIMINISHING_RETURNS_WINDOW: 5 iterations < 0.5% each → warn user and suggest stopping
22	STATE_DIR: .experiments/state/<run-id>/ (timestamped dir per run — see .claude/rules/artifact-lifecycle.md)
23	SENTINEL_SLUG_FORMULA: \|
24	eval "$(bash "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/git_slugs.sh")"
25	# Sentinel path: ${TMPDIR:-/tmp}/claude-commit-auth-${REPO_SLUG}-${BRANCH_SLUG} # tmpdir-exempt: user-shell-boundary
26	# Bash state is lost between tool calls — re-source git_slugs.sh at each use site; it is the only authorized slug form.
27	```
28
29	<!-- Note: STATE_DIR (.experiments/state/) holds per-iteration artifacts (diary, experiments.jsonl).
30	Hypothesis pipeline outputs (hypotheses.jsonl, checkpoint.json, journal.md) go to .experiments/<run-id>/ (RUN_DIR).
31	These are two separate directories by design — see protocol.md for layout. -->
32
33	Agent strategy mapping (`agent_strategy` in config → ideation agent to spawn):
34
35	\| `agent_strategy` \| Specialist agent \| When to use \|
36	\| --- \| --- \| --- \|
37	\| `auto` \| heuristic \| Default — infer from metric_cmd keywords \|
38	\| `perf` \| `foundry:perf-optimizer` \| latency, throughput, memory, GPU utilization \|
39	\| `code` \| `foundry:sw-engineer` \| coverage, complexity, lines, coupling \|
40	\| `ml` \| `research:scientist` \| accuracy, loss, F1, AUC, BLEU \|
41	\| `arch` \| `foundry:solution-architect` \| coupling, cohesion, modularity metrics \|
42
43	Auto-inference keyword heuristics (when `agent_strategy: auto` or omitted; checked against `## Goal` text AND metric command):
44
45	Precedence order (first match wins; ML keywords beat test-framework keywords). ML-specific compound terms (not bare tokens) required — prevents over-triggering on `eval`/`train`/`val` as common words:
46	- contains `accuracy`, `loss` (paired with `train_loss`/`val_loss`/`eval_loss`), `f1_score`, `auc_roc`, `auroc`, `train_step`, `val_acc`, `eval_loss`, `epoch`, `gradient`, `tensor`, `overfit`, `generaliz`, `regulariz`, `validation`, `dropout`, `weight_decay`, `lr_schedule`, `cross_val`, `precision`, `recall`, OR explicit `--scientist` flag → `ml` → `research:scientist`
47	- contains `time`, `latency`, `bench`, `throughput`, `memory` → `perf` → `foundry:perf-optimizer`
48	- contains `pytest`, `coverage`, `complexity` → `code` → `foundry:sw-engineer`
49	- no keyword match → `perf` (default fallback) — WARN: print `⚠ No keyword match — defaulting to 'perf' strategy. If this is an ML task, set agent_strategy: ml in program.md.` Log resolved agent + reason in state.json `strategy_resolution`.
50
51	Bare tokens `eval`, `train`, `val` (without compound suffix) do NOT trigger `ml` routing — too common in non-ML contexts (test eval scripts, training-environment configs, validator command names).
52
53	Stuck escalation sequence (at STUCK_THRESHOLD consecutive discards):
54
55	1. Switch agent type. Rotation by current strategy:
56
57	\| Current strategy \| Next strategy \| Escalation agent \|
58	\| --- \| --- \| --- \|
59	\| `code` \| `ml` \| `research:scientist` \|
60	\| `ml` \| `perf` \| `foundry:perf-optimizer` \|
61	\| `perf` \| `code` \| `foundry:sw-engineer` \|
62	\| `arch` \| `code` \| `foundry:sw-engineer` (fallback `foundry:solution-architect` if sw-engineer unavailable) \|
63	\| `auto` \| infer from resolved strategy \| follow rotation row for whichever concrete strategy `auto` heuristics resolved t

People who liked this also viewed

6 picks

Type