fix(agents): seed claude-cli fallback prompts with prior-session context (#69973) by stainlu · Pull Request #72069 · openclaw/openclaw

stainlu · 2026-04-26T08:45:57Z

Why

When a claude-cli attempt fails with a fallbackable error (e.g. 402 billing limit), the next candidate runs cold. Claude Code keeps full session history in ~/.claude/projects/<id>/<sessionId>.jsonl, but the fallback runner only sees what OpenClaw assembles from its own transcript — and that transcript is empty for claude-cli sessions because OpenClaw doesn't mirror Claude's local JSONL turns into its own session file.

@steipete's triage: "the next candidate only sees what OpenClaw assembles from its own transcript/context. If the OpenClaw transcript is truncated or missing the prior CLI-backed turns, the fallback candidate starts cold while Claude later resumes fine. Concrete fix: when a claude-cli attempt fails with a fallbackable error, seed the fallback prompt from the OpenClaw transcript/compacted summary before trying the non-CLI candidate."

Pattern

Mirrors what Claude Code itself does on resume after /compact: prefer the explicit summary, then append the most recent post-boundary turns up to a char budget. Inferred from the upstream JSONL shape (type: "summary", type: "system" subtype: "compact_boundary" with compactMetadata) and confirmed against the published session-management behavior. References:

Last-N raw replay was rejected as the wrong shape: it diverges from how Claude Code itself replays after compaction, misses context if /compact already happened (the older turns are gone, only the summary remains), and chews more prompt budget than the equivalent summary.

What changed

File	Role
`src/gateway/cli-session-history.claude.ts`	New `readClaudeCliFallbackSeed` walks the JSONL with awareness of `type: "summary"` and `compact_boundary` entries. Pre-boundary turns are dropped (they're now in the summary). Multiple compactions: latest summary wins.
`src/gateway/cli-session-history.ts`	Re-export the new reader + type.
`src/agents/command/attempt-execution.helpers.ts`	New `formatClaudeCliFallbackPrelude` / `buildClaudeCliFallbackContextPrelude`. Tool blocks coalesce to compact `(tool call: …)` / `(tool result: …)` hints. Newest turns kept when truncating. Summary clearly labeled `(truncated)` if it overflows. `resolveFallbackRetryPrompt` gains an optional `priorContextPrelude`.
`src/agents/command/attempt-execution.ts`	`runAgentAttempt` builds the prelude when `isFallbackRetry && providerOverride !== claude-cli && cliSessionBindings.claude-cli.sessionId is set`. Same-provider fallbacks (claude-cli → claude-cli) are unaffected — Claude's own `--resume` already works there.

Verified

33 tests in src/agents/command/attempt-execution.test.ts (12 added)
12 tests in src/gateway/cli-session-history.test.ts (7 added)
pnpm tsgo:core + pnpm tsgo:core:test clean
pnpm format + pnpm lint clean on touched files
Regression-catching verified: removing the prelude prepend in resolveFallbackRetryPrompt fails both new prelude cases, restoring the cold-start behavior — confirming the test pins the behavior (not just the helper signature).

Risks flagged

Privacy: same surface as the existing augmentChatHistoryWithCliSessionImports — we already read user's Claude JSONL for display. No new ground.
Path safety: reuses existing resolveClaudeCliSessionFilePath validation, which already rejects ..-style escapes.
Summary quality: if Claude's own /compact produced a malformed summary (see #46602), we inherit that. Mitigation: the prelude is clearly labeled "Prior session context (from claude-cli)" so the fallback model treats it as a hint, not authoritative state.
Char budget: defaults to 8k chars; truncates at word boundaries with … so the model never sees a mid-summary fragment.

Out of scope

Codex CLI sessions (openai-codex as primary). Claude is the driver of the user-pain reports here. The same mechanism could be extended provider-by-provider as those CLIs expose their own structured transcripts.
Two-way: when a non-CLI session falls back to claude-cli. Claude has --resume and a session id of its own; cross-pollination there is a separate problem.

Files

src/gateway/cli-session-history.claude.ts:302+ — parser extension
src/gateway/cli-session-history.ts:11+ — re-export
src/agents/command/attempt-execution.helpers.ts:104+ — formatter + builder + extended resolveFallbackRetryPrompt
src/agents/command/attempt-execution.ts:265+ — integration in runAgentAttempt

greptile-apps · 2026-04-26T08:49:37Z

Greptile Summary

This PR seeds fallback retry prompts with prior claude-cli session context by parsing Claude Code's local JSONL (summary + post-boundary turns) and prepending a labeled prelude to the fallback candidate's prompt, mirroring the resume-after-compaction shape Claude Code uses itself. The implementation is clean, well-tested, and the integration guard (isFallbackRetry && !isClaudeCliProvider(providerOverride)) correctly scopes the change to cross-provider fallbacks only.

Confidence Score: 4/5

Safe to merge; only P2 edge-case findings, no critical logic or security issues.

All findings are P2. The main concern is that a summary JSONL entry without a following compact_boundary (e.g. crash mid-compaction) leaves pre-summary turns in windowedTurns, causing the fallback model to see redundant context. This is a degenerate edge case and the impact is mild (extra context, not wrong context). The break-on-first-oversized-turn behavior is intentional by design. Test coverage is comprehensive (7 new cases in the gateway suite, 12 in the helpers suite) and the regression-catching verification was explicitly confirmed by the author.

src/gateway/cli-session-history.claude.ts — the summary/compact_boundary interaction in readClaudeCliFallbackSeed

Comments Outside Diff (1)

src/gateway/cli-session-history.claude.ts, line 567-574 (link)

summary entry without a following compact_boundary surfaces pre-summary turns as recentTurns

When a type: "summary" entry appears in the JSONL but is not followed by a compact_boundary (e.g., a crash mid-compaction, or a future Claude Code variant), processing the summary only updates summaryText — it does not reset windowedTurns. All turns accumulated before the summary remain in the window and end up in recentTurns, so the fallback model receives both the summary and the turns it is based on, doubling that portion of the context.

The compact_boundary handler does reset the window, so the normal flow (summary → compact_boundary) is correct. The gap is the orphaned-summary case. A simple guard would close it:

const explicitSummary = extractSummaryText(parsed);
if (explicitSummary) {
  summaryText = explicitSummary;
  // Drop turns that are now represented by this summary so they are not
  // double-counted if no compact_boundary follows.
  windowedTurns = [];
  toolNameRegistry.clear();
  continue;
}

Prompt To Fix With AI

This is a comment left during a code review.
Path: src/gateway/cli-session-history.claude.ts
Line: 567-574

Comment:
**`summary` entry without a following `compact_boundary` surfaces pre-summary turns as `recentTurns`**

When a `type: "summary"` entry appears in the JSONL but is not followed by a `compact_boundary` (e.g., a crash mid-compaction, or a future Claude Code variant), processing the `summary` only updates `summaryText` — it does **not** reset `windowedTurns`. All turns accumulated before the summary remain in the window and end up in `recentTurns`, so the fallback model receives both the summary and the turns it is based on, doubling that portion of the context.

The `compact_boundary` handler does reset the window, so the normal flow (`summary → compact_boundary`) is correct. The gap is the orphaned-summary case. A simple guard would close it:

```typescript
const explicitSummary = extractSummaryText(parsed);
if (explicitSummary) {
  summaryText = explicitSummary;
  // Drop turns that are now represented by this summary so they are not
  // double-counted if no compact_boundary follows.
  windowedTurns = [];
  toolNameRegistry.clear();
  continue;
}
```

How can I resolve this? If you propose a fix, please make it concise.

Prompt To Fix All With AI

This is a comment left during a code review.
Path: src/gateway/cli-session-history.claude.ts
Line: 567-574

Comment:
**`summary` entry without a following `compact_boundary` surfaces pre-summary turns as `recentTurns`**

When a `type: "summary"` entry appears in the JSONL but is not followed by a `compact_boundary` (e.g., a crash mid-compaction, or a future Claude Code variant), processing the `summary` only updates `summaryText` — it does **not** reset `windowedTurns`. All turns accumulated before the summary remain in the window and end up in `recentTurns`, so the fallback model receives both the summary and the turns it is based on, doubling that portion of the context.

The `compact_boundary` handler does reset the window, so the normal flow (`summary → compact_boundary`) is correct. The gap is the orphaned-summary case. A simple guard would close it:

```typescript
const explicitSummary = extractSummaryText(parsed);
if (explicitSummary) {
  summaryText = explicitSummary;
  // Drop turns that are now represented by this summary so they are not
  // double-counted if no compact_boundary follows.
  windowedTurns = [];
  toolNameRegistry.clear();
  continue;
}
```

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: src/agents/command/attempt-execution.helpers.ts
Line: 120-128

Comment:
**`break` on first oversized turn silently drops all older turns**

The loop walks newest → oldest and breaks as soon as one turn does not fit the remaining budget. This means that if the second-newest turn is large (e.g. a long tool result), any older — possibly much shorter — turns are silently omitted, even though they would fit individually.

For a "most-recent contiguous window" semantic this is intentional and the comment describes it as such. If a sparse/non-contiguous window would give better context coverage, a `continue` instead of `break` would achieve that. Since the PR explicitly chose contiguous semantics this is flagged as an observation rather than a blocking issue, but it's worth a conscious decision given the char budget is already quite generous at 8 k.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "fix(agents): seed claude-cli fallback pr..." | Re-trigger Greptile}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: add644115b

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

stainlu · 2026-04-26T14:03:27Z

Addressed the review on 9987e7797f:

Codex P1 (provider gating): plumb originalProvider (the user-requested provider for the run) into runAgentAttempt. The seed now requires isClaudeCliProvider(originalProvider) so a stale cliSessionBindings["claude-cli"] from an unrelated past run cannot bleed into a fallback chain that started on another provider.
Codex P2 (summary/boundary pairing): restructured readClaudeCliFallbackSeed to queue summaries into a pendingSummary slot and flush each compact_boundary's pair atomically. A boundary with no preceding summary now correctly falls back to the boundary's own content rather than serving an older summary alongside fresh post-boundary turns. New regression test in cli-session-history.test.ts pins this; trailing summary without a boundary is also covered.
Greptile P2 (break behavior): kept the newest-first break semantic (a contiguous "what was happening just before the failure" window beats sparse coverage with gaps) and documented the design choice inline so a future maintainer doesn't reflexively flip it to skip-and-continue.

Verified: 33 attempt-execution + 14 cli-session-history tests; broader src/agents/command suite green (63/63); pnpm tsgo:core + type-aware oxlint clean on touched files.

clawsweeper · 2026-04-26T14:10:49Z

Codex automated review: keeping this open.

Keep this PR open. Current main still does not implement the PR’s intended cross-provider fallback seed from Claude CLI session history, while the live PR remains open, unmerged, focused on the linked fallback handoff bug, and now includes review-followup commits for source-provider gating and compact-boundary pairing.

Best possible solution:

Keep this PR open for maintainer review. The best path is to land this PR or an equivalent implementation that seeds a bounded Claude CLI summary/recent-turn prelude only when the fallback chain started on claude-cli and retries on another provider, with the regression coverage already proposed in src/agents/command and src/gateway.

What I checked:

current main retry helper lacks prior context seed: resolveFallbackRetryPrompt only accepts body, isFallbackRetry, and sessionHasHistory; when sessionHasHistory is false it returns the original body, which is the cold fallback case this PR targets. (src/agents/command/attempt-execution.helpers.ts:104, fb40ed99a7f0)
current main attempt construction does not read Claude session history: runAgentAttempt builds effectivePrompt only from the request body, fallback flag, and OpenClaw session-history boolean. There is no Claude CLI fallback seed reader call or prelude passed to the prompt helper. (src/agents/command/attempt-execution.ts:262, fb40ed99a7f0)
current main fallback caller lacks source-provider plumbing: The fallback loop passes candidate provider/model and isFallbackRetry into runAgentAttempt, but does not pass the original requested provider. The PR adds that field so Claude-only seeding can be gated by the failed source provider rather than stale session bindings. (src/agents/agent-command.ts:954, fb40ed99a7f0)
current Claude JSONL support is display/import only: Current main exports readClaudeCliSessionMessages and uses augmentChatHistoryWithCliSessionImports; there is no readClaudeCliFallbackSeed, ClaudeCliFallbackSeed, or summary/compact-boundary fallback seed API. (src/gateway/cli-session-history.claude.ts:302, fb40ed99a7f0)
current tests still pin no-history retry as cold: Current helper tests assert fallback retries with no OpenClaw session history preserve the original body, and the gateway history tests cover chat-history imports rather than fallback seed extraction from Claude summaries/recent turns. (src/agents/command/attempt-execution.test.ts:34, fb40ed99a7f0)
linked issue discussion supports the PR’s problem statement: The linked bug discussion includes triage identifying this as a fallback handoff bug and recommending that a failed claude-cli attempt seed the non-CLI fallback candidate with prior context. The issue was later closed based on broader CLI transcript persistence, but that did not add this PR’s Claude JSONL fallback seed path to current main.

Remaining risk / open question:

Closing this PR now would drop an active, focused implementation candidate for a fallback continuity gap that current main still lacks.
Before landing, maintainers should still review the privacy/context-transfer boundary: the PR intentionally reads Claude Code’s local JSONL and sends a bounded summary/recent-turn prelude to a different fallback provider.
The PR’s JSONL parsing relies on Claude Code’s observed summary and compact-boundary record shape; the added tests cover known edge cases, but maintainers should decide whether that dependency contract is acceptable.

Codex Review notes: model gpt-5.5, reasoning high; reviewed against fb40ed99a7f0.

…ext (openclaw#69973) When a claude-cli attempt failed with a fallbackable error (e.g. a 402 billing limit), the next candidate -- typically a non-CLI provider -- ran with no prior conversation context. Claude Code keeps its own JSONL session under ~/.claude/projects/, but the fallback runner only sees what OpenClaw assembles from its own transcript, which is empty for claude-cli sessions. The fallback model therefore behaved as if the conversation just started, even though Claude later resumed fine. Resolution mirrors what Claude Code itself does on resume after compaction: prefer the explicit `/compact` summary, then append the most recent post-boundary turns up to a char budget. Concretely: - `readClaudeCliFallbackSeed` (gateway): walks the Claude JSONL with awareness of `type: "summary"` and `type: "system", subtype: "compact_boundary"` entries. Pre-boundary turns are dropped (they are represented by the summary); post-boundary turns become the recent-window. Multiple compactions are handled by preferring the latest summary. Path safety reuses the existing `resolveClaudeCliSessionFilePath` validation. - `formatClaudeCliFallbackPrelude` / `buildClaudeCliFallbackContext\ Prelude` (agents helpers): format the harvested seed into a labeled prelude. Tool blocks are coalesced to compact "(tool call: name)" / "(tool result: …)" hints to keep the prompt budget honest. Newest turns are kept first when truncating; the summary is clearly labeled "(truncated)" if it overflows. - `resolveFallbackRetryPrompt`: gains an optional `priorContextPrelude` that prepends before the existing retry marker. Empty/whitespace preludes are ignored; first-attempt prompts are unchanged. - `runAgentAttempt`: builds the prelude when `isFallbackRetry === true` AND the new candidate is non-claude-cli AND a Claude-cli session binding is present. Same-provider fallbacks (claude-cli to claude-cli) are unaffected because Claude's own --resume still works. Verified the new tests (12 in cli-session-history, 12 added to attempt-execution) catch the regression: removing the prelude prepend in resolveFallbackRetryPrompt makes both new prelude cases fail, restoring the original cold-start behavior. References: - https://code.claude.com/docs/en/how-claude-code-works - "Inside Claude Code: The Session File Format" https://databunny.medium.com/inside-claude-code-the-session-file-format-and-how-to-inspect-it-b9998e66d56b

…formatter Local default oxlint did not run --type-aware so the warning was missed on the initial commit; CI surfaced it via check-lint. Hoist the heading into a named const so its length is read directly without the assertion.

…ndary Addresses review on openclaw#72069: - Codex P1 ("Gate Claude prelude seeding by source provider"): the guard checked the *current* fallback candidate but not the failed attempt. A session that still carried a stale cliSessionBindings["claude-cli"] from an unrelated past run would inject Claude transcript context into a fallback chain that started on a different provider (e.g. openai -> openai-codex), leaking irrelevant prior conversation. Plumb `originalProvider` (the user-requested provider for the chain) through to runAgentAttempt and require `isClaudeCliProvider(originalProvider)` before reading Claude history. - Codex P2 ("Prefer latest compact boundary when summary is missing"): the resolver always preferred the most recent explicit summary, so a later compaction without its own summary entry (rare crash case) paired stale summary text with post-latest-boundary turns. Restructure readClaudeCliFallbackSeed to queue summaries into pendingSummary and flush each boundary's pair atomically. A boundary with no preceding summary now correctly falls back to the boundary's own content rather than serving an older summary alongside fresh turns. - Greptile P2 (newest-first break vs sparse coverage): the formatFallbackTurns walk intentionally stops on the first oversized turn so the prelude stays a contiguous "what was happening just before the failure" window. Document the design choice inline so a future maintainer doesn't reflexively change it to skip-and-continue. Tests: - New gateway cases for the boundary-without-summary edge case and for trailing summaries written without a paired boundary. - existing 33 attempt-execution + 14 cli-session-history tests still pass; broader src/agents/command suite stays green (63/63).

The required-typed param introduced in 9987e7797f broke attempt-execution.cli.test.ts and auth-profile-runtime-contract.test.ts which construct runAgentAttempt params without an originalProvider field. Make it optional and explicitly require the typeof check before passing to isClaudeCliProvider so a missing field correctly skips the seed (defensive default for fallback paths that didn't plumb the original provider through, no-op for non-fallback paths).

obviyus

Verified the claude-cli fallback handoff: when the primary claude-cli attempt fails, the fallback prompt now carries Claude Code session context instead of starting cold.

Maintainer follow-up: made the source provider explicit, trimmed review comments, added contiguous-window coverage, and added the Unreleased changelog entry.

Local gate: pnpm test src/agents/command/attempt-execution.test.ts src/agents/command/attempt-execution.cli.test.ts src/gateway/cli-session-history.test.ts src/agents/auth-profile-runtime-contract.test.ts; pnpm tsgo:core; pnpm tsgo:test:src.

…ndary Addresses review on #72069: - Codex P1 ("Gate Claude prelude seeding by source provider"): the guard checked the *current* fallback candidate but not the failed attempt. A session that still carried a stale cliSessionBindings["claude-cli"] from an unrelated past run would inject Claude transcript context into a fallback chain that started on a different provider (e.g. openai -> openai-codex), leaking irrelevant prior conversation. Plumb `originalProvider` (the user-requested provider for the chain) through to runAgentAttempt and require `isClaudeCliProvider(originalProvider)` before reading Claude history. - Codex P2 ("Prefer latest compact boundary when summary is missing"): the resolver always preferred the most recent explicit summary, so a later compaction without its own summary entry (rare crash case) paired stale summary text with post-latest-boundary turns. Restructure readClaudeCliFallbackSeed to queue summaries into pendingSummary and flush each boundary's pair atomically. A boundary with no preceding summary now correctly falls back to the boundary's own content rather than serving an older summary alongside fresh turns. - Greptile P2 (newest-first break vs sparse coverage): the formatFallbackTurns walk intentionally stops on the first oversized turn so the prelude stays a contiguous "what was happening just before the failure" window. Document the design choice inline so a future maintainer doesn't reflexively change it to skip-and-continue. Tests: - New gateway cases for the boundary-without-summary edge case and for trailing summaries written without a paired boundary. - existing 33 attempt-execution + 14 cli-session-history tests still pass; broader src/agents/command suite stays green (63/63).

obviyus · 2026-04-28T08:06:15Z

Landed on main.

Landed source commit: aac8e06
Merge commit: 8da2fb1

Thanks @stainlu.

…ndary Addresses review on openclaw#72069: - Codex P1 ("Gate Claude prelude seeding by source provider"): the guard checked the *current* fallback candidate but not the failed attempt. A session that still carried a stale cliSessionBindings["claude-cli"] from an unrelated past run would inject Claude transcript context into a fallback chain that started on a different provider (e.g. openai -> openai-codex), leaking irrelevant prior conversation. Plumb `originalProvider` (the user-requested provider for the chain) through to runAgentAttempt and require `isClaudeCliProvider(originalProvider)` before reading Claude history. - Codex P2 ("Prefer latest compact boundary when summary is missing"): the resolver always preferred the most recent explicit summary, so a later compaction without its own summary entry (rare crash case) paired stale summary text with post-latest-boundary turns. Restructure readClaudeCliFallbackSeed to queue summaries into pendingSummary and flush each boundary's pair atomically. A boundary with no preceding summary now correctly falls back to the boundary's own content rather than serving an older summary alongside fresh turns. - Greptile P2 (newest-first break vs sparse coverage): the formatFallbackTurns walk intentionally stops on the first oversized turn so the prelude stays a contiguous "what was happening just before the failure" window. Document the design choice inline so a future maintainer doesn't reflexively change it to skip-and-continue. Tests: - New gateway cases for the boundary-without-summary edge case and for trailing summaries written without a paired boundary. - existing 33 attempt-execution + 14 cli-session-history tests still pass; broader src/agents/command suite stays green (63/63).

openclaw-barnacle Bot added gateway Gateway runtime agents Agent runtime and tooling size: L labels Apr 26, 2026

greptile-apps Bot reviewed Apr 26, 2026

View reviewed changes

Comment thread src/agents/command/attempt-execution.helpers.ts

chatgpt-codex-connector Bot reviewed Apr 26, 2026

View reviewed changes

Comment thread src/agents/command/attempt-execution.ts Outdated

Comment thread src/gateway/cli-session-history.claude.ts Outdated

stainlu force-pushed the fix/issue-69973-cli-fallback-context-seed branch from 5510d33 to 12d8e44 Compare April 28, 2026 04:39

stainlu requested a review from a team as a code owner April 28, 2026 04:39

openclaw-barnacle Bot added docs Improvements or additions to documentation cli CLI command changes scripts Repository scripts labels Apr 28, 2026

openclaw-barnacle Bot added size: XL and removed size: L labels Apr 28, 2026

stainlu force-pushed the fix/issue-69973-cli-fallback-context-seed branch from 26dc4ad to e06f042 Compare April 28, 2026 06:46

openclaw-barnacle Bot added size: L and removed docs Improvements or additions to documentation cli CLI command changes scripts Repository scripts size: XL labels Apr 28, 2026

obviyus force-pushed the fix/issue-69973-cli-fallback-context-seed branch from e06f042 to 23d1e50 Compare April 28, 2026 07:57

obviyus requested a review from a team as a code owner April 28, 2026 07:57

obviyus self-assigned this Apr 28, 2026

stainlu added 2 commits April 28, 2026 13:31

stainlu and others added 4 commits April 28, 2026 13:31

fix(agents): require claude fallback source provider

54c2e99

fix: seed claude-cli fallback context (openclaw#72069) (thanks @stainlu)

aac8e06

obviyus force-pushed the fix/issue-69973-cli-fallback-context-seed branch from 23d1e50 to aac8e06 Compare April 28, 2026 08:04

obviyus approved these changes Apr 28, 2026

View reviewed changes

obviyus merged commit 8da2fb1 into openclaw:main Apr 28, 2026
58 of 61 checks passed

clawsweeper Bot mentioned this pull request Apr 30, 2026

Fallback retry rewrites user prompt into synthetic transcript message (should preserve original body) #61006

Closed

ogt-redknie pushed a commit to ogt-redknie/OPENX that referenced this pull request May 2, 2026

fix: seed claude-cli fallback context (openclaw#72069) (thanks @stainlu)

b065ef2

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 9, 2026

fix: seed claude-cli fallback context (openclaw#72069) (thanks @stainlu)

9aa107b

clawsweeper Bot mentioned this pull request May 11, 2026

CLI runner: add orphaned-user-message repair for fallback retries #54192

Open

globalcaos pushed a commit to globalcaos/tinkerclaw that referenced this pull request May 13, 2026

fix: seed claude-cli fallback context (openclaw#72069) (thanks @stainlu)

c32312e

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 24, 2026

fix: seed claude-cli fallback context (openclaw#72069) (thanks @stainlu)

18a9284

jameslcowan pushed a commit to jameslcowan/openclaw that referenced this pull request Jun 2, 2026

fix: seed claude-cli fallback context (openclaw#72069) (thanks @stainlu)

0cdead8

sablehead pushed a commit to sablehead/openclaw that referenced this pull request Jun 10, 2026

fix: seed claude-cli fallback context (openclaw#72069) (thanks @stainlu)

26a3197

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(agents): seed claude-cli fallback prompts with prior-session context (#69973)#72069

fix(agents): seed claude-cli fallback prompts with prior-session context (#69973)#72069
obviyus merged 6 commits into
openclaw:mainfrom
stainlu:fix/issue-69973-cli-fallback-context-seed

stainlu commented Apr 26, 2026

Uh oh!

greptile-apps Bot commented Apr 26, 2026 •

edited

Loading

Comments Outside Diff (1)

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

Uh oh!

stainlu commented Apr 26, 2026

Uh oh!

clawsweeper Bot commented Apr 26, 2026 •

edited

Loading

Uh oh!

obviyus left a comment

Uh oh!

Uh oh!

obviyus commented Apr 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

stainlu commented Apr 26, 2026

Why

Pattern

What changed

Verified

Risks flagged

Out of scope

Files

Uh oh!

greptile-apps Bot commented Apr 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 4/5

Comments Outside Diff (1)

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

stainlu commented Apr 26, 2026

Uh oh!

clawsweeper Bot commented Apr 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

obviyus left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

obviyus commented Apr 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

greptile-apps Bot commented Apr 26, 2026 •

edited

Loading

clawsweeper Bot commented Apr 26, 2026 •

edited

Loading