perf(task): prune validation and implicit scheduler graphs

[morluto]

Closes #76.

Prune unused task validation work and replace dense graphs with predecessor chains for wholly implicit sequential tasks. Mixed/explicit graphs retain their existing behavior. Both initial scheduling and retry rearming use the reduced representation.

Workload	Before	After	Speedup
Validate 3,000 implicit steps	100.629 ms	0.242 ms	416×
Validate 500 self-retrying steps	734.573 ms	0.051 ms	~14,500×
Build 3,000-step scheduler graph	224.843 ms	0.872 ms	258×

The graph uses 2,999 edges instead of 4,498,500. Peak traced allocations: 39.1 MB → 1.32 MB. Measurements are validation/graph microbenchmarks on macOS arm64, Python 3.12.13, against 602d730, with five timing samples. They do not measure complete task execution; earlier-target retries and mixed graphs retain their existing validation costs.

Reproduce:

PYTHONPATH=src:packages/agentenv-protocol/src .venv/bin/python tst/benchmarks/task_graphs.py --base 602d730 --steps 1000 3000 --retry-steps 500 --repeats 5

Validation:

• The baseline now loads its own methods, dependency helpers and state schema; regression coverage prevents mixed revisions and constructor mismatches.

• 157 focused task/retry tests passed, including actual execution, persisted retries, resume/end boundaries, callbacks and cancellation.

• Full unit and protocol suites: 6,238 passed, 13 skipped.

• Deterministic regressions check graph edge counts and absence of unused prefix slicing/self-retry ancestry.

• Independent review completed with no actionable findings; plugin API check and diff checks passed.

Separate commits keep validation pruning and scheduler graph reduction reviewable. This PR is independent of the journal, Explorer and existing context-diff PRs.

Confidence Score: 5/5

The PR appears safe to merge; both earlier benchmark findings are addressed, and no new issue was established.

Summary

Task validation and scheduling now skip unnecessary work for tasks that use the default sequential order. The PR also adds a benchmark that compares these costs with methods loaded from a chosen Git revision.

• Implicit tasks skip unused checks while they are being validated.
• Wholly implicit tasks keep their order with a predecessor chain.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    A[Build scheduler state] --> B{All active steps implicit?}
    B -->|Yes| C[Link each step to its predecessor]
    B -->|No| D[Keep the existing dependency edges]
    C --> E[Use dependency map for retry rearming]
    D --> E

Loading

_{Reviews (3) · Last reviewed commit: "fix(benchmark): load task baselines from..." · Reviewed by Greptile}

[earakely-scale]

Nice PR, approved