Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Strategy: Dependency Constraint · yogsoth-ai | DecimalAI

yogsoth-ai/dependency-constraint

SkillScore82Maintained

What must be completed first? — Dependency chain analysis + prerequisite graph construction

productivityby yogsoth-ai

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Project Management pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/yogsoth-ai-dependency-constraint/SKILL.md

decimalai skills pull yogsoth-ai-dependency-constraint --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/yogsoth-ai-dependency-constraint/SKILL.md https://app.decimal.ai/s/yogsoth-ai-dependency-constraint/SKILL.md

Raw file: /s/yogsoth-ai-dependency-constraint/SKILL.md ↗ · machine-readable summary at /s/yogsoth-ai-dependency-constraint.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 7d ago

—+10% tokens0% turns82% of whole cases pass with skill (18/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-02	✗→✓	▲ Improved	-4%	0%
case-03	✗→✓	▲ Improved	-4%	0%
case-11	✗→✓	▲ Improved	-4%	0%
case-12	✗→✓	▲ Improved	-36%	0%
case-13	✗→✓	▲ Improved	-47%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

yogsoth-ai / de-anthropocentric-research-engine

skills/dependency-constraint/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 8/3/2026Version v1

Skill contents

SKILL.md

Strategy: Dependency Constraint

Methodology

Systematic dependency analysis:

Task decomposition: Break experiment into atomic tasks
Dependency typing: Finish-to-Start, Start-to-Start, Finish-to-Finish, Start-to-Finish
Graph construction: Build directed acyclic graph (DAG)
Critical path: Longest path through the DAG
Critical chain: Critical path + resource contention
Prerequisite identification: External dependencies, approvals, data availability

Dependency categories: | Category | Examples | |----------|----------| | Technical | Code A needs library B, model needs data | | Sequential | Train before evaluate, design before implement | | Resource | Same GPU needed by two tasks | | External | API access, dataset release, paper review | | Knowledge | Need result X to decide approach Y |

Execution Flow

Build Dependency Graph → call dependency-graph-construction SOP

Input: task list, known dependencies
Output: DAG with typed edges

Identify Critical Chain → call critical-chain-identification SOP

Input: DAG + resource assignments
Output: critical chain with contention points

Rank Sensitivity → invoke sensitivity-ranking tactic

Determine which dependencies are most binding

Report → synthesize dependency assessment

Critical path length
Binding dependency constraints
Parallelization opportunities

Budget Gate

| Resource | Budget | Notes | |----------|--------|-------| | Subagent calls | ≤5 | 2 SOPs + synthesis | | Iterations | ≤2 | Re-build if tasks change | | Output size | ≤3000 tokens | Graph summary + critical chain |

Available Tactics

Optional, no fixed order; the final leaf is always a sop.

| Tactic | When to use | | --- | --- | | sensitivity-ranking | Rank constraints by sensitivity — which ones most impact the outcome if they shift |

Available SOPs

Optional, no fixed order; the final leaf is always a sop.

| SOP | When to use | | --- | --- | | critical-chain-identification | Identify the critical chain — longest path considering resource contention | | dependency-graph-construction | Build task dependency graph with predecessor/successor relationships |

Want to see if yogsoth-ai/dependency-constraint would activate on your phrasing? Try it before you install — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/22/2026, 13:48 UTC · 22 cases

Measured on 20 comparable cases — below the 22 needed to publish a lift, so this result is shown here but not on the registry card or leaderboard.

Lift

+32%

82% vs 50%

Agent turns

0%

1.0 vs 1.0

Tokens

+10%

2,041.5 vs 1,861.3

fail → pass · 9pass → pass · 9fail → fail · 2pass → fail · 2

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 We are running a complex ML experiment involving custom dataset scraping, annotation verification, pre-training a transformer model, fine-tuning on downstream tasks, and evaluating against baseline benchmarks. Some tasks share the same 4-GPU cluster, and fine-tuning depends on annotated data releases. Can you map out all technical, resource, and sequential prerequisites, determine our critical bottleneck path under GPU constraints, and output a structured assessment showing key constraints and potential parallel execution tracks?	fail→fail	24,428	11,584	-53%	1	1	0%	3,681	1,096	-70%	0	0	—
case-02 I need to coordinate a multi-stage software benchmarking experiment across five team sub-projects. We have shared API access limits, code refactoring prerequisites, and downstream evaluation steps that rely on intermediate test results. Please analyze these project milestones and produce a dependency report identifying the primary critical chain under resource constraints, sensitive bottleneck dependencies, and opportunities to run tasks in parallel.	fail→pass	20,781	19,670	-5%	1	1	0%	3,668	3,505	-4%	0	0	—
case-03 Our team is planning an AI safety evaluation suite where paper review approvals, safety alignment dataset downloads, model fine-tuning, and human annotation must occur in sequence or concurrently where possible. We have tight GPU memory and human evaluator bandwidth limits. Please generate a full dependency analysis mapping predecessor relationships, identifying binding resource constraints along the critical execution path, and listing achievable parallelization options.	fail→pass	19,664	15,279	-22%	1	1	0%	3,321	3,172	-4%	0	0	—
case-04 Our engineering team has 10 user stories for the upcoming 2-week iteration, with story point estimates totaling 34 points. Our historical velocity is 28 points per iteration. Which stories should we commit to during sprint planning to optimize team capacity without overcommitting?	pass→fail	12,274	16,888	+38%	1	1	0%	2,103	3,091	+47%	0	0	—
case-05 We are purchasing cloud compute instances for a 6-month research grant. On-demand A100 instances cost $3.09/hr, reserved instances cost $1.80/hr with a 1-year commitment, and spot instances average $0.90/hr with interruption risks. Calculate the expected financial cost across these purchasing options for 5,000 GPU hours.	pass→pass	17,859	26,966	+51%	1	1	0%	3,361	5,322	+58%	0	0	—
case-06 Provide a pull request review checklist for testing python code formatting, docstring compliance, unit test coverage thresholds, and static type checking in CI/CD pipelines.	pass→pass	14,561	14,528	-0%	1	1	0%	2,339	2,824	+21%	0	0	—
case-07 We are scheduling a pipeline where model training requires clean dataset generation to complete before starting, while log streaming must start simultaneously with inference testing. Rather than standard generic arrows, how should these predecessor relationships be categorized in a formal task graph?	pass→fail	12,850	18,560	+44%	1	1	0%	2,243	1,892	-16%	0	0	—
case-08 We have two independent tasks, Model Pretraining A (40 hours) and Synthetic Data Generation B (35 hours). Normally Task A is the longest path. However, both tasks require 100% of GPU Cluster 1, which only has capacity to run one task at a time. What concept accounts for this resource constraint beyond simple topological path length?	pass→pass	9,284	6,611	-29%	1	1	0%	1,490	1,632	+10%	0	0	—
case-09 In an NLP benchmarking initiative, human subject ethics approval delays initial data gathering, while a limited 8-card GPU server restricts simultaneous fine-tuning runs. Categorize these two specific bottlenecks into standard dependency categories.	pass→pass	7,299	9,651	+32%	1	1	0%	1,231	2,034	+65%	0	0	—
case-10 After building a task graph for our vision pipeline, we have 12 potential dependencies. We need to identify which specific dependency constraint will cause the largest shift in project completion time if delayed. Which analysis step should be executed?	fail→fail	8,057	3,523	-56%	1	1	0%	1,324	1,110	-16%	0	0	—
case-11 When analyzing a complex multi-task AI deployment project from scratch using standard structured workflow procedures, what primary analysis steps should be performed and in what sequence?	fail→pass	15,573	19,102	+23%	1	1	0%	2,503	2,407	-4%	0	0	—
case-12 We are configuring an automated subagent orchestration plan to conduct dependency analysis on an enterprise software rollout. What budget limit governs the total allowed subagent calls for this workflow?	fail→pass	8,436	2,024	-76%	1	1	0%	1,263	804	-36%	0	0	—
case-13 When generating the final dependency assessment synthesis report for an AI research pipeline, what max output size threshold must be enforced for token budget compliance?	fail→pass	9,839	1,905	-81%	1	1	0%	1,512	799	-47%	0	0	—
case-14 In our RL training campaign, we cannot choose our reward shaping algorithm parameters for Task Y until the hyperparameter sweep results from Task X are finalized. What category of dependency is this?	fail→pass	5,857	5,433	-7%	1	1	0%	933	1,379	+48%	0	0	—
case-15 In a distributed benchmark experiment, continuous integration telemetry collection must complete at the exact same time as the final integration benchmark run. What dependency type represents this constraint?	pass→pass	4,506	5,654	+25%	1	1	0%	784	1,409	+80%	0	0	—
case-16 During a data center migration for ML serving, the legacy inference API cannot be shut down until the new v2 inference API service successfully starts up. What formal dependency relationship applies here?	pass→pass	4,720	8,752	+85%	1	1	0%	819	1,999	+144%	0	0	—
case-17 Before constructing a directed acyclic graph (DAG) for a large-scale evaluation experiment, what is the required initial systematic analysis step?	pass→pass	8,200	3,856	-53%	1	1	0%	1,228	1,172	-5%	0	0	—
case-18 What key analytical sections must be included in the final dependency assessment synthesis report for an ML platform migration?	fail→pass	16,940	11,572	-32%	1	1	0%	2,686	2,304	-14%	0	0	—
case-19 If tasks or underlying assumptions change during dependency mapping, how many iteration passes are permitted under budget limits?	fail→pass	12,369	1,798	-85%	1	1	0%	1,797	763	-58%	0	0	—
case-20 We have two project risks: (1) PyTorch model compilation requires CUDA 12 support, and (2) release of the target test set relies on an external research lab's public repository publish date. How are these categorized?	pass→pass	7,385	5,288	-28%	1	1	0%	1,277	1,370	+7%	0	0	—
case-21 In a clinical AI model deployment, system architecture design must take place prior to writing code, and model training must take place prior to validation testing. What dependency category governs these operational ordering constraints?	fail→pass	7,405	4,486	-39%	1	1	0%	1,176	1,229	+5%	0	0	—
case-22 When evaluating a task dependency DAG without considering physical hardware or resource limitations, how is the critical path identified?	pass→pass	13,056	11,702	-10%	1	1	0%	2,411	2,505	+4%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +32 percentage points is the difference between those two pass rates over the 20 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 50%
Pass rate with the skill: 81.8%
Cases attempted: 22
Cases counted toward lift: 20
Cases that regressed: 2
Skill version measured: v1
Run completed: 8/22/2026, 13:48 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-02

Prompt

I need to coordinate a multi-stage software benchmarking experiment across five team sub-projects. We have shared API access limits, code refactoring prerequisites, and downstream evaluation steps that rely on intermediate test results. Please analyze these project milestones and produce a dependency report identifying the primary critical chain under resource constraints, sensitive bottleneck dependencies, and opportunities to run tasks in parallel.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run82

Live evalNeeds ≥20 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+36% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 6 days ago

Other measured skills in the registry, with their headline benchmark lift.