Add extraction quality signals for context assimilation

[densmirnov]

Task: 202607031338-KHMQAV
Title: Add extraction quality signals for context assimilation
Canonical task record: .agentplane/tasks/202607031338-KHMQAV/README.md

Summary

Add extraction quality signals for context assimilation

Add a derived extraction quality report that flags over-fragmentation, unsupported summaries, and risky normalization before wiki publication.

Scope

• In scope: Add a derived extraction quality report that flags over-fragmentation, unsupported summaries, and risky normalization before wiki publication.
• Out of scope: unrelated refactors not required for "Add extraction quality signals for context assimilation".

Verification

• State: ok
• Note: Extraction quality report implemented and focused context checks passed.
• Canonical workflow state lives in the task README.

Raw evidence

• Updated: 2026-07-03T13:38:55.819Z
• Branch: task/202607031338-KHMQAV/add-extraction-quality-signals-for-context-assim
• Head: computed live by agentplane pr check / agentplane integrate

 .../commands/context/extraction-apply.unit.test.ts |  90 ++++++++++
 .../agentplane/src/commands/context/extraction.ts  |   1 +
 .../agentplane/src/context/extraction-writer.ts    | 182 +++++++++++++++++++++
 3 files changed, 273 insertions(+)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9d00a6f11e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Honor span refs in precision quality checks

When a valid SGR item carries precision through span_refs (or a coverage span_id) but its source_refs only name the source file, this filter still classifies it as source_precision_gap because it only inspects SgrSourceRef fields. That makes the new quality report warn on span-addressed extractions even though the warning text says span-addressed refs are acceptable, so maximum-assimilation tasks using derived source spans get false rework signals. Consider treating item span refs as precise too.

Useful? React with 👍 / 👎.