explore-run

Name: explore-run
Author: lllllllama

161kinstalls10forks

How to install explore-run?

lllllllama/rigorpilot-skills/explore-run

$npx -y skills add lllllllama/RigorPilot-Skills --skill explore-run

Installs into the current project.

Summary

Rigor Improve / Rigor Explore run leaf skill for bounded exploratory evidence in deep learning research repositories. Use when the researcher explicitly authorizes exploratory runs such as small-subset validation, short-cycle guess-and-check, batch sweeps, idle-GPU search, or qui

Files · 1

View on GitHub

SKILL.md

1	# explore-run
2
3	Use this as the Rigor Improve / Rigor Explore run leaf skill. The installed slug
4	remains `explore-run` for compatibility.
5
6	Use the shared operating principles in
7	`../../references/agent-operating-principles.md`; this skill should guide
8	candidate run planning while preserving model judgment about the active repo.
9
10	## When to apply
11
12	- When the researcher explicitly authorizes exploratory runs.
13	- When the task is a small-subset validation, short-cycle training probe, batch sweep, idle-GPU search, or quick transfer-learning trial.
14	- When the output should rank candidate runs rather than certify trusted success.
15
16	## When not to apply
17
18	- When the user wants trusted training execution or conservative verification.
19	- When there is no explicit exploratory authorization.
20	- When the task is repository setup, intake, or debugging.
21
22	## Clear boundaries
23
24	- This skill owns exploratory execution planning and summary only.
25	- Use `ai-research-explore` instead when the task spans both current_research coordination and exploratory code changes.
26	- It may hand off actual command execution to `minimal-run-and-audit` or `run-train`.
27	- It should keep experiment state isolated from the trusted baseline.
28	- It should prefer small-subset and short-cycle checks before heavier exploratory runs.
29	- It should label run results as bounded evidence and explain when a comparison
30	is not directly fair.
31
32	## Ranking Semantics
33
34	- Pre-execution candidate selection uses three factors: `cost`, `success_rate`, and `expected_gain`.
35	- Default weights should stay conservative unless the researcher explicitly provides `selection_weights`.
36	- Budget pruning still applies after scoring through `max_variants` and `max_short_cycle_runs`.
37	- If runs are executed later, downstream ranking should switch to real execution evidence, not stay purely heuristic.
38
39	## Variant Spec Hints
40
41	- Use `variant_axes` to define the candidate dimension grid.
42	- Use `subset_sizes` and `short_run_steps` to express exploratory run scale.
43	- Use `selection_weights` to rebalance `cost`, `success_rate`, and `expected_gain`.
44	- Use `primary_metric` and `metric_goal` so downstream ranking can order executed candidates consistently.
45
46	## Output expectations
47
48	- `explore_outputs/CHANGESET.md`
49	- `explore_outputs/SCIENTIFIC_CHANGELOG.md`
50	- `explore_outputs/COMPARABILITY_REPORT.md`
51	- `explore_outputs/TOP_RUNS.md`
52	- `explore_outputs/status.json`
53
54	## Notes
55
56	Use `references/execution-policy.md`, `../../references/explore-variant-spec.md`, `../../references/deep-learning-experiment-principles.md`, `scripts/plan_variants.py`, and `scripts/write_outputs.py`.

People who liked this also viewed

6 picks

Type

explore-run

161kinstalls10forks

View on GitHub

How to install explore-run?

lllllllama/rigorpilot-skills/explore-run

$npx -y skills add lllllllama/RigorPilot-Skills --skill explore-run

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# explore-run
2
3	Use this as the Rigor Improve / Rigor Explore run leaf skill. The installed slug
4	remains `explore-run` for compatibility.
5
6	Use the shared operating principles in
7	`../../references/agent-operating-principles.md`; this skill should guide
8	candidate run planning while preserving model judgment about the active repo.
9
10	## When to apply
11
12	- When the researcher explicitly authorizes exploratory runs.
13	- When the task is a small-subset validation, short-cycle training probe, batch sweep, idle-GPU search, or quick transfer-learning trial.
14	- When the output should rank candidate runs rather than certify trusted success.
15
16	## When not to apply
17
18	- When the user wants trusted training execution or conservative verification.
19	- When there is no explicit exploratory authorization.
20	- When the task is repository setup, intake, or debugging.
21
22	## Clear boundaries
23
24	- This skill owns exploratory execution planning and summary only.
25	- Use `ai-research-explore` instead when the task spans both current_research coordination and exploratory code changes.
26	- It may hand off actual command execution to `minimal-run-and-audit` or `run-train`.
27	- It should keep experiment state isolated from the trusted baseline.
28	- It should prefer small-subset and short-cycle checks before heavier exploratory runs.
29	- It should label run results as bounded evidence and explain when a comparison
30	is not directly fair.
31
32	## Ranking Semantics
33
34	- Pre-execution candidate selection uses three factors: `cost`, `success_rate`, and `expected_gain`.
35	- Default weights should stay conservative unless the researcher explicitly provides `selection_weights`.
36	- Budget pruning still applies after scoring through `max_variants` and `max_short_cycle_runs`.
37	- If runs are executed later, downstream ranking should switch to real execution evidence, not stay purely heuristic.
38
39	## Variant Spec Hints
40
41	- Use `variant_axes` to define the candidate dimension grid.
42	- Use `subset_sizes` and `short_run_steps` to express exploratory run scale.
43	- Use `selection_weights` to rebalance `cost`, `success_rate`, and `expected_gain`.
44	- Use `primary_metric` and `metric_goal` so downstream ranking can order executed candidates consistently.
45
46	## Output expectations
47
48	- `explore_outputs/CHANGESET.md`
49	- `explore_outputs/SCIENTIFIC_CHANGELOG.md`
50	- `explore_outputs/COMPARABILITY_REPORT.md`
51	- `explore_outputs/TOP_RUNS.md`
52	- `explore_outputs/status.json`
53
54	## Notes
55
56	Use `references/execution-policy.md`, `../../references/explore-variant-spec.md`, `../../references/deep-learning-experiment-principles.md`, `scripts/plan_variants.py`, and `scripts/write_outputs.py`.

People who liked this also viewed

6 picks

Type