Skip to content

fix(session): skill suitability block isolation and regression tests (#1504) - #1532

Open
santoshkumarradha wants to merge 2 commits into
devfrom
fix/1504-skill-turn
Open

santoshkumarradha wants to merge 2 commits into
devfrom
fix/1504-skill-turn

Conversation

@santoshkumarradha

Copy link
Copy Markdown
Member

Summary

Closes #1504: Skill suitability block leaks into visible transcript.

Findings & Changes

  • The core isolation logic (skill suitability block rides the model request via user.said while the store/journal retains only the person's words, and the dim skills carried line is the visible UI channel) already landed on origin/dev via Hosted chat says it stopped while the run keeps running #1345.
  • Added end-to-end regression test in internal/session/skillturn_display_test.go verifying that the store transcript holds only user words while the model request receives the block.
  • Updated documentation in internal/manual/chat/skills-a-turn-used.md.
  • Added unreleased changelog entry in docs/changes/unreleased/1504-transcript-keeps-your-words.md.

Pre-existing Failure Root Cause & Resolution

  • Root-caused and resolved the two pre-existing internal/session package test failures:
    1. hears() test leak: lane-news sighting was leaking across tests; isolated the state.
    2. Ignored-folder receipt test: corrected to match the canonical path formatting.
  • go test ./internal/session/... now runs with 0 failures (down from 2 on base origin/dev).

—
Drafted with CodeAF · reviewed and owned by the author

…skills (#1504)

The skills block rides the copy the model reads; the record the surfaces
draw keeps the person's own words, and the dim skills-carried line is the
one visible channel. Adds the end-to-end regression test through the real
chat door, and lands the full internal/session suite green: the lane-news
test helper no longer hands an earlier test's held sighting to the first
reader of the next, and the ignored-folder receipt test wants the
canonical spelling the receipt actually prints.
@santoshkumarradha santoshkumarradha added this to the Reliable agent milestone Sep 26, 2026
@santoshkumarradha santoshkumarradha added bug Something the code does that it should not area:session The engine — turns, tasks, the toolbelt, checkpoints sev:papercut A wording, a hint, a small wrongness that costs a moment labels Sep 26, 2026
…tore back

The end-to-end test read brain.Messages straight after the turn's events
closed, but the transcript reaches the store through the batching chat
log: the read raced the writer and on CI lost, finding an empty thread.
Close the log first, the way chatlog_test reads its thread back.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area:session The engine — turns, tasks, the toolbelt, checkpoints bug Something the code does that it should not sev:papercut A wording, a hint, a small wrongness that costs a moment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Skill-suitability block leaks into the visible transcript — keep it model-context only

2 participants