edge-pipeline-orchestrator

Name: edge-pipeline-orchestrator
Author: mphinance

15stars3forks

How to install edge-pipeline-orchestrator?

mphinance/alpha-skills/edge-pipeline-orchestrator

$npx -y skills add mphinance/alpha-skills --skill edge-pipeline-orchestrator

Installs into the current project.

Summary

Orchestrate the full edge research pipeline from candidate detection through strategy design, review, revision, and export. Use when coordinating multi-stage edge research workflows end-to-end.

Files · 1

View on GitHub

SKILL.md

1	# Edge Pipeline Orchestrator
2
3	Coordinate all edge research stages into a single automated pipeline run.
4
5	## When to Use
6
7	- Run the full edge pipeline from tickets (or OHLCV) to exported strategies
8	- Resume a partially completed pipeline from the drafts stage
9	- Review and revise existing strategy drafts with feedback loop
10	- Dry-run the pipeline to preview results without exporting
11
12	## Workflow
13
14	1. Load pipeline configuration from CLI arguments
15	2. Run auto_detect stage if --from-ohlcv is provided (generates tickets from raw OHLCV data)
16	3. Run hints stage to extract edge hints from market summary and anomalies
17	4. Run concepts stage to synthesize abstract edge concepts from tickets and hints
18	5. Run drafts stage to design strategy drafts from concepts
19	6. Run review-revision feedback loop:
20	- Review all drafts (max 2 iterations)
21	- PASS verdicts accumulated; REJECT verdicts accumulated
22	- REVISE verdicts trigger apply_revisions and re-review
23	- Remaining REVISE after max iterations downgraded to research_probe
24	7. Export eligible drafts (PASS + export_ready_v1 + exportable entry_family)
25	8. Write pipeline_run_manifest.json with full execution trace
26
27	## CLI Usage
28
29	```bash
30	# Full pipeline from tickets
31	python3 scripts/orchestrate_edge_pipeline.py \
32	--tickets-dir path/to/tickets/ \
33	--output-dir reports/edge_pipeline/
34
35	# Full pipeline from OHLCV
36	python3 scripts/orchestrate_edge_pipeline.py \
37	--from-ohlcv path/to/ohlcv.csv \
38	--output-dir reports/edge_pipeline/
39
40	# Resume from drafts stage
41	python3 scripts/orchestrate_edge_pipeline.py \
42	--resume-from drafts \
43	--drafts-dir path/to/drafts/ \
44	--output-dir reports/edge_pipeline/
45
46	# Review-only mode
47	python3 scripts/orchestrate_edge_pipeline.py \
48	--review-only \
49	--drafts-dir path/to/drafts/ \
50	--output-dir reports/edge_pipeline/
51
52	# Dry run (no export)
53	python3 scripts/orchestrate_edge_pipeline.py \
54	--tickets-dir path/to/tickets/ \
55	--output-dir reports/edge_pipeline/ \
56	--dry-run
57	```
58
59	## Output
60
61	All artifacts are written to `--output-dir`:
62
63	```
64	output-dir/
65	├── pipeline_run_manifest.json
66	├── tickets/ (from auto_detect)
67	├── hints/hints.yaml (from hints)
68	├── concepts/edge_concepts.yaml
69	├── drafts/*.yaml
70	├── exportable_tickets/*.yaml
71	├── reviews_iter_0/*.yaml
72	├── reviews_iter_1/*.yaml (if needed)
73	└── strategies/<candidate_id>/
74	├── strategy.yaml
75	└── metadata.json
76	```
77
78	## Claude Code LLM-Augmented Workflow
79
80	Run the LLM-augmented pipeline entirely within Claude Code:
81
82	1. Run auto_detect to produce `market_summary.json` + `anomalies.json`
83	2. Claude Code analyzes data and generates edge hints
84	3. Save hints to a YAML file:
85
86	```yaml
87	- title: Sector rotation into industrials
88	observation: Tech underperforming while industrials show relative strength
89	symbols: [CAT, DE, GE]
90	regime_bias: Neutral
91	mechanism_tag: flow
92	preferred_entry_family: pivot_breakout
93	hypothesis_type: sector_x_stock
94	```
95
96	4. Run orchestrator with `--llm-ideas-file` and `--promote-hints`:
97
98	```bash
99	python3 scripts/orchestrate_edge_pipeline.py \
100	--tickets-dir path/to/tickets/ \
101	--llm-ideas-file llm_hints.yaml \
102	--promote-hints \
103	--as-of 2026-02-28 \
104	--max-synthetic-ratio 1.5 \
105	--strict-export \
106	--output-dir reports/edge_pipeline/
107	```
108
109	### Optional Flags
110
111	- `--as-of YYYY-MM-DD` — forwarded to hints stage for date filtering
112	- `--strict-export` — export-eligible drafts with any warn finding get REVISE instead of PASS
113	- `--max-synthetic-ratio N` — cap synthetic tickets to N × real ticket count (floor: 3)
114	- `--overlap-threshold F` — condition overlap threshold for concept deduplication (default: 0.75)
115	- `--no-dedup` — disable concept deduplication
116
117	Note: `--llm-ideas-file` and `--promote-hints` are effective only during full pipeline runs.
118	`--resume-from drafts` and `--review-only` skip hints/concepts stages, so these flags are ignored.
119
120	## Resources
121
122	- `references/pipeline_flow.md` — Pipeline stages, data contracts, and architecture
123	- `references/revision_loop_rules.md` — Review-revision feedback loop rules and heuristics

People who liked this also viewed

6 picks

Type

edge-pipeline-orchestrator

15stars3forks

View on GitHub

How to install edge-pipeline-orchestrator?

mphinance/alpha-skills/edge-pipeline-orchestrator

$npx -y skills add mphinance/alpha-skills --skill edge-pipeline-orchestrator

Installs into the current project.

Summary

Orchestrate the full edge research pipeline from candidate detection through strategy design, review, revision, and export. Use when coordinating multi-stage edge research workflows end-to-end.

Files · 1

View on GitHub

SKILL.md

1	# Edge Pipeline Orchestrator
2
3	Coordinate all edge research stages into a single automated pipeline run.
4
5	## When to Use
6
7	- Run the full edge pipeline from tickets (or OHLCV) to exported strategies
8	- Resume a partially completed pipeline from the drafts stage
9	- Review and revise existing strategy drafts with feedback loop
10	- Dry-run the pipeline to preview results without exporting
11
12	## Workflow
13
14	1. Load pipeline configuration from CLI arguments
15	2. Run auto_detect stage if --from-ohlcv is provided (generates tickets from raw OHLCV data)
16	3. Run hints stage to extract edge hints from market summary and anomalies
17	4. Run concepts stage to synthesize abstract edge concepts from tickets and hints
18	5. Run drafts stage to design strategy drafts from concepts
19	6. Run review-revision feedback loop:
20	- Review all drafts (max 2 iterations)
21	- PASS verdicts accumulated; REJECT verdicts accumulated
22	- REVISE verdicts trigger apply_revisions and re-review
23	- Remaining REVISE after max iterations downgraded to research_probe
24	7. Export eligible drafts (PASS + export_ready_v1 + exportable entry_family)
25	8. Write pipeline_run_manifest.json with full execution trace
26
27	## CLI Usage
28
29	```bash
30	# Full pipeline from tickets
31	python3 scripts/orchestrate_edge_pipeline.py \
32	--tickets-dir path/to/tickets/ \
33	--output-dir reports/edge_pipeline/
34
35	# Full pipeline from OHLCV
36	python3 scripts/orchestrate_edge_pipeline.py \
37	--from-ohlcv path/to/ohlcv.csv \
38	--output-dir reports/edge_pipeline/
39
40	# Resume from drafts stage
41	python3 scripts/orchestrate_edge_pipeline.py \
42	--resume-from drafts \
43	--drafts-dir path/to/drafts/ \
44	--output-dir reports/edge_pipeline/
45
46	# Review-only mode
47	python3 scripts/orchestrate_edge_pipeline.py \
48	--review-only \
49	--drafts-dir path/to/drafts/ \
50	--output-dir reports/edge_pipeline/
51
52	# Dry run (no export)
53	python3 scripts/orchestrate_edge_pipeline.py \
54	--tickets-dir path/to/tickets/ \
55	--output-dir reports/edge_pipeline/ \
56	--dry-run
57	```
58
59	## Output
60
61	All artifacts are written to `--output-dir`:
62
63	```
64	output-dir/
65	├── pipeline_run_manifest.json
66	├── tickets/ (from auto_detect)
67	├── hints/hints.yaml (from hints)
68	├── concepts/edge_concepts.yaml
69	├── drafts/*.yaml
70	├── exportable_tickets/*.yaml
71	├── reviews_iter_0/*.yaml
72	├── reviews_iter_1/*.yaml (if needed)
73	└── strategies/<candidate_id>/
74	├── strategy.yaml
75	└── metadata.json
76	```
77
78	## Claude Code LLM-Augmented Workflow
79
80	Run the LLM-augmented pipeline entirely within Claude Code:
81
82	1. Run auto_detect to produce `market_summary.json` + `anomalies.json`
83	2. Claude Code analyzes data and generates edge hints
84	3. Save hints to a YAML file:
85
86	```yaml
87	- title: Sector rotation into industrials
88	observation: Tech underperforming while industrials show relative strength
89	symbols: [CAT, DE, GE]
90	regime_bias: Neutral
91	mechanism_tag: flow
92	preferred_entry_family: pivot_breakout
93	hypothesis_type: sector_x_stock
94	```
95
96	4. Run orchestrator with `--llm-ideas-file` and `--promote-hints`:
97
98	```bash
99	python3 scripts/orchestrate_edge_pipeline.py \
100	--tickets-dir path/to/tickets/ \
101	--llm-ideas-file llm_hints.yaml \
102	--promote-hints \
103	--as-of 2026-02-28 \
104	--max-synthetic-ratio 1.5 \
105	--strict-export \
106	--output-dir reports/edge_pipeline/
107	```
108
109	### Optional Flags
110
111	- `--as-of YYYY-MM-DD` — forwarded to hints stage for date filtering
112	- `--strict-export` — export-eligible drafts with any warn finding get REVISE instead of PASS
113	- `--max-synthetic-ratio N` — cap synthetic tickets to N × real ticket count (floor: 3)
114	- `--overlap-threshold F` — condition overlap threshold for concept deduplication (default: 0.75)
115	- `--no-dedup` — disable concept deduplication
116
117	Note: `--llm-ideas-file` and `--promote-hints` are effective only during full pipeline runs.
118	`--resume-from drafts` and `--review-only` skip hints/concepts stages, so these flags are ignored.
119
120	## Resources
121
122	- `references/pipeline_flow.md` — Pipeline stages, data contracts, and architecture
123	- `references/revision_loop_rules.md` — Review-revision feedback loop rules and heuristics

People who liked this also viewed

6 picks

Type