Forumics logo Forumics
Issue 01

Back to issue

All Cline known issues

CLAIM C130 Supported 2026-11-15

GitHub issue #15483 in cline/cline, opened 2026-11-01 by Lohnegrim, records a report of two related problems with run_commands, both measured on disk rather than inferred from chat text: (A) a tool call whose input carries an unexpected envelope is rejected with a bare Invalid input that never names the field, while the same validator does name fields for a different error class; (B) context compaction removes the examples that keep the call form correct and locks in the broken form — before the first compaction the session had made 378 run_commands calls, all of them in the correct form, and had never produced the envelope at all, while the compacted prefix retained 10 tool-call examples including 5 run_commands, all five in the broken envelope form, each followed by its retained is_error true result, and zero correct run_commands examples; after the first burst the session produced 17 consecutive rejections and never recovered on its own. System cited: Cline 4.1.26, VS Code 1.144.0, Windows 10.0.26300, provider openai-compatible via vLLM.

Verbatim Excerpts from Cited Records
“**(A) A tool call whose input carries an unexpected envelope is rejected with a bare `Invalid input` that never names the field.**”
— Lohnegrim · cline/cline (issue 15483 opening body, What happened section) · 2026-11-01T13:15:42Z ·GitHub Issue ↗
“No field name, no expected shape, no hint that `arguments` is the intruder.”
— Lohnegrim · cline/cline (issue 15483 opening body, What happened section) · 2026-11-01T13:15:42Z ·GitHub Issue ↗
“The two forms are perfectly separated: the envelope never once worked.”
— Lohnegrim · cline/cline (issue 15483 opening body, What happened section) · 2026-11-01T13:15:42Z ·GitHub Issue ↗
“Before the first compaction the session had made **378 `run_commands` calls, all of them in the correct form, and had never produced the envelope at all.**”
— Lohnegrim · cline/cline (issue 15483 opening body, What happened section) · 2026-11-01T13:15:42Z ·GitHub Issue ↗
Counter-record: Issue open as of the fetch; all 4 comments read. linear-code[bot] (2026-11-01) linked the report to CLINE-3719. John-Rood (2026-11-01) analyzed the retained examples as recent, not representative: findCutIndex walks the transcript from the end accumulating estimated tokens to the preserveRecentTokens default of 20,000, reading neither is_error nor the tool name, so the retained block can hold five failures and none of the 378 successes; he traced the sidecar file that survives restarts and reaches the next summary, supplied a compaction-examples-audit script to prune the rejected pairs, verified the sidecar hash binds only the transcript side so pruning is safe, and proposed a retention-layer fix of reserving the most recent successful pair per enabled tool; he confirmed the (A) read and that surfacing issue.keys in the prettifier is the whole fix. GreenTruck50 (2026-11-14) said he is disabling autocompaction until it is fixed. John-Rood (2026-11-14) replied that disabling compaction stops the poisoned projection but the sidecar stays on disk and any later compaction starts from it, and that pruning with the audit script removes the bad pairs while keeping the summary. No maintainer fix and no retraction of the report appear in the thread. Related published claims C048 (#13437, same tool and same bare Invalid input, different malformation) and C093 (#13346, run_commands output handling) are distinct reports; this claim documents only #15483. Falsification probe: If GitHub issue #15483 is deleted without archive, if the author edits the opening body so the quoted sentences no longer match verbatim, or if the author posts a retraction of the report, this claim drops to Withdrawn. Audit verification: Verified 100% continuous verbatim substring against official GitHub REST API by scripts/verify_claim_verbatim.py.