weichen 03b4fdabaf Restrict the source tag to material reads; never tag the model's own files ...

Why: The first E4 live comparison regressed badly (toolFailures 10->24,
toolCalls 63->166) and the instance transcript shows the loop: tagging
kernel read_file/grep results put a trailing [src_N] line on the
model's own workspace files, the model treated it as external
tampering, and re-read/re-edit cycles cascaded into the read-before-
write and file-changed guards.

Why: The ledger only ever needed SOURCES - searches always, and
workspace reads only under inputs/ (the upload material area). Grep
cannot distinguish materials from self-produced files, so it stays
untagged.

Evidence: go test ./... green; new anchor TestEvidenceLedger
OnlyTagsMaterialReads pins the inputs/-only rule; the re-run live
comparison follows on the Dev side.

Impact: evidenceTools loses read_file/grep; isEvidenceRead gates
workspace reads to inputs/; everything else (search tags, escalation
reminders, prompts) unchanged.

Refs: Dev task feature/agent-e4-escalation-ledger (fix-up of #131).

2026-10-10 18:11:24 +00:00

..

testdata

Check final replies before the direct run stops

2026-10-01 13:54:15 +00:00

compose_test.go

Keep artifact ids and byte counts out of the final reply

2026-10-02 14:02:19 +00:00

compose.go

Repair the quality-gate failure chain observed on real runs

2026-10-09 12:34:21 +00:00

context_window_test.go

Keep original materials and cleared results reachable across turns

2026-10-09 14:28:39 +00:00

deliverable_latest_test.go

Keep tool-result dumps out of the deliverable gate and cards

2026-10-09 14:56:34 +00:00

dual_speed_nudge_test.go

Strip inline think blocks from subagent summaries and card bodies

2026-10-10 10:01:15 +00:00

e4_escalation_test.go

Restrict the source tag to material reads; never tag the model's own files

2026-10-10 18:11:24 +00:00

factcheck_test.go

Add R4 escalation reminders and the structured source-id evidence ledger (E4)

2026-10-10 17:41:38 +00:00

factcheck.go

Restrict the source tag to material reads; never tag the model's own files

2026-10-10 18:11:24 +00:00

final_reply_test.go

Keep the final reply clean when checks continue or the verifier is skipped

2026-10-10 07:48:03 +00:00

hostbridge.go

Route conversations through a squad entry when selected

2026-10-01 09:34:53 +00:00

llm_direct_e1_test.go

Enable parallel tool execution, fix delegate truncation, sandbox deps layer

2026-10-10 09:02:10 +00:00

llm_direct_n3_test.go

Make the kernel transcript the authoritative cross-turn context

2026-10-09 15:38:33 +00:00

llm_direct.go

Restrict the source tag to material reads; never tag the model's own files

2026-10-10 18:11:24 +00:00

llm_test.go

Add R4 escalation reminders and the structured source-id evidence ledger (E4)

2026-10-10 17:41:38 +00:00

llm_think_test.go

Return deliverable generation to the agent and fix the artifact-card root cause

2026-10-01 03:32:30 +00:00

llm_verify_test.go

Keep the final reply clean when checks continue or the verifier is skipped

2026-10-10 07:48:03 +00:00

llm.go

Enable parallel tool execution, fix delegate truncation, sandbox deps layer

2026-10-10 09:02:10 +00:00

mock.go

Make background subagents observable as the slow lane of dual-speed tasks

2026-10-10 06:40:00 +00:00

model_test.go

Rewind tracked files and fall back before the first token

2026-10-03 17:31:56 +00:00

model.go

Store secrets apart from settings and assemble context at the end

2026-10-04 18:20:10 +00:00

permission_ask_test.go

Validate plugin agent declarations and wire the squad ask port

2026-10-07 05:14:06 +00:00

planless_nudge_test.go

Add the working-method base prompt and demote the compose rewrite

2026-10-09 23:12:53 +00:00

review_test.go

Let subagents run in the background and be reviewed

2026-10-01 15:29:01 +00:00

review.go

Let subagents run in the background and be reviewed

2026-10-01 15:29:01 +00:00

runner.go

Validate plugin agent declarations and wire the squad ask port

2026-10-07 05:14:06 +00:00

squad_runner_test.go

Route conversations through a squad entry when selected

2026-10-01 09:34:53 +00:00

squad_runner.go

Enable parallel tool execution, fix delegate truncation, sandbox deps layer

2026-10-10 09:02:10 +00:00

sub_loop_test.go

Stream tool progress and drop hard agent limits

2026-10-02 02:17:02 +00:00

sub_loop.go

Enable parallel tool execution, fix delegate truncation, sandbox deps layer

2026-10-10 09:02:10 +00:00

toolselect_test.go

Cap tool results by character and gate stale file writes

2026-10-03 16:21:02 +00:00

transcript_compact.go

Make the kernel transcript the authoritative cross-turn context

2026-10-09 15:38:33 +00:00

transcript_seed_test.go

Make the kernel transcript the authoritative cross-turn context

2026-10-09 15:38:33 +00:00

transcript_seed.go

Make the kernel transcript the authoritative cross-turn context

2026-10-09 15:38:33 +00:00

usage_n3_test.go

Split usage accounting end to end, retry by default, and take the WP6 pins (N3 WP6)

2026-10-07 17:17:56 +00:00

verify.go

Add R4 escalation reminders and the structured source-id evidence ledger (E4)

2026-10-10 17:41:38 +00:00

workflow_compiled_test.go

Keep original materials and cleared results reachable across turns

2026-10-09 14:28:39 +00:00

workflow_compiled.go

Enable parallel tool execution, fix delegate truncation, sandbox deps layer

2026-10-10 09:02:10 +00:00

workspace_tools_test.go

Open the kernel workspace tools and the markdown-source build pipeline

2026-10-09 18:01:06 +00:00

wp5_ab_test.go

Keep the final reply clean when checks continue or the verifier is skipped

2026-10-10 07:48:03 +00:00