[deep-report] Add max_turns + deterministic pre-steps to Documentation Unbloat and Package Specification Extractor (overkill candidates)

[github-actions]

Description

Two workflows consistently appear in the "overkill" portfolio quadrant, consuming high tokens for tasks with significant deterministic components:

Workflow	Turns	Tokens/run	Issue
Documentation Unbloat	58 (failed run)	4.8M	New failure Apr 27, also 34.8min runtime
Package Specification Extractor	51	3.0M	resource_heavy + partially_reducible

Both have agentic_fraction ≈ 0.50 — meaning roughly half their turns are data-gathering operations that could be done deterministically before the agent starts.

Concrete fixes (from Optimization Kit #28742 Prompt 4):

For Package Specification Extractor:

1. Add bash pre-step: list packages from go.mod, fetch existing specs, pre-load top-10 changed packages from git log
2. Reduce agentic phase to enrichment-only (not discovery)
3. Add max-turns: 25

For Documentation Unbloat:

1. Add max-turns: 35 (current 58 = no guard, caused 34.8min failed run)
2. Replace Playwright-based content fetching with direct curl/gh api calls where possible
3. Add timeout guard on external content loading

Expected Impact

Documentation Unbloat: prevent future 34.8min wasted runs. Package Extractor: reduce 3M → ~1.5M tokens/run. Combined weekly saving: ~4.5M tokens.

Suggested Agent

Agentic Optimization Kit

Estimated Effort

Medium (1–4 hours)

Data Source

DeepReport Intelligence Briefing — April 27, 2026 (run §25003521428); Weekly Workflow Analysis #28687; Agentic Optimization Kit #28742

Generated by DeepReport - Intelligence Gathering Agent · ● 473.9K · ◷

• expires on Apr 29, 2026, 3:35 PM UTC

[github-actions]

This issue was automatically closed because it expired on 2026-04-29T15:35:40.908Z.

Closed by Workflow