Files
openclaw/qa/scenarios/runtime/tools/fs-read.yaml
Dallin Romney c1c231ced9 fix(qa): reject runtime tool intent without result evidence (#113502)
* fix(qa): require runtime tool result evidence

* fix(qa): preserve meaningful tool coverage gates

* refactor(qa): colocate runtime parity projection

* fix(qa): keep passing tool totals evidence-backed
2026-07-27 11:00:59 +08:00

52 lines
1.8 KiB
YAML

title: "Runtime tool fixture — fs.read"
scenario:
id: runtime-tool-fs-read
surface: runtime-tools
runtimePairLane: core
coverage:
secondary:
- agent-runtime.tool-fs-read
objective: Verify file read behavior is tracked across OpenClaw and Codex while Codex owns read natively.
successCriteria:
- OpenClaw may expose OpenClaw read while Codex app-server mode may omit duplicate OpenClaw dynamic read.
- Mock provider read plans are reported as fixture intent, not as actual runtime tool calls.
- The row stays report-only until failure-path injection proves native Codex read behavior directly.
docsRefs:
- qa/scenarios/index.yaml
codeRefs:
- src/agents/agent-tools.read.ts
- extensions/qa-lab/src/runtime-tool-fixture.ts
execution:
kind: flow
summary: Exercise the read runtime tool family.
config:
toolName: read
toolCoverage:
family: fs.read
actualTool: read
bucket: codex-native-workspace
expectedLayer: codex-native-workspace
required: false
tracking: "#80312"
codexDefaultImpact: P4
qaImpact: P2
action: model native read behavior separately from provider-plan capture
reason: Codex app-server intentionally owns read natively; QA mock failure-path capture currently reports provider-plan args, not proven Codex native read behavior.
knownHarnessGap:
issue: "#80312"
reason: QA mock failure-path capture currently reports provider-plan args, not proven Codex native read behavior.
promptSnippet: "target=read"
failurePromptSnippet: "failure target=read"
flow:
steps:
- name: exercises fs.read happy and failure paths
actions:
- call: runRuntimeToolFixture
saveAs: result
args:
- ref: env
- ref: config
detailsExpr: result