Repository navigation
Conversation
Adds rf-bdd for behaviour-driven specifications in native .robot files: house rules, Gherkin mapping, step vocabularies, scenario state, a Robocop configuration, and two runnable example projects (resource and World-library strategies), verified on RF 7.4.2 and 7.5.1 and on Robocop 8.2.10 and 9.4. - rf-language points to rf-bdd; its casing gotcha follows the project's Robocop configuration; a comma-separated --select selects nothing (it is not rejected). - The context hook and rf-test-architect route BDD work to rf-bdd. - Evals: the sut-bdd-library fixture (base, steps, review variants), a BDD grader with tests and golden solutions, five tasks and a trigger set. The preflight selects tasks by fixture root. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FHEmzL3ErAhGQykKWDx2wP
Skill Evaluation: rf-mobile (narrow, treatment, N=3, z-ai/glm-5.3-flash)Model: OpenRouter spend
Gaterf-skill-eval reportRuns: 3 · pass: 3 · fail: 0 · incomplete: 0 · skipped checks: 0 · estimated cost (OpenRouter list prices): $0.02 Pass rate = runs whose gate result is Tasks
Skills
Skipped / errored checksNone. Incomplete runsNone. Process checks (treatment arm, not in deltas)
Adversarial tier (reported only, never gates)None. |
Skill Evaluation: rf-desktop (narrow, treatment, N=3, z-ai/glm-5.3-flash)Model: OpenRouter spend
Gaterf-skill-eval reportRuns: 3 · pass: 2 · fail: 1 · incomplete: 0 · skipped checks: 0 · estimated cost (OpenRouter list prices): $0.03 · turn-limit stops: 1 Pass rate = runs whose gate result is Tasks
Skills
Skipped / errored checksNone. Incomplete runsNone. Process checks (treatment arm, not in deltas)
Adversarial tier (reported only, never gates)None. |
Skill Evaluation: rf-web (narrow, treatment, N=3, z-ai/glm-5.3-flash)Model: OpenRouter spend
Gaterf-skill-eval reportRuns: 6 · pass: 6 · fail: 0 · incomplete: 0 · skipped checks: 0 · estimated cost (OpenRouter list prices): $0.05 Pass rate = runs whose gate result is Tasks
Skills
Skipped / errored checksNone. Incomplete runsNone. Process checks (treatment arm, not in deltas)
Adversarial tier (reported only, never gates)None. |
Skill Evaluation: rf-api (narrow, treatment, N=3, z-ai/glm-5.3-flash)Model: OpenRouter spend
Gaterf-skill-eval reportRuns: 6 · pass: 6 · fail: 0 · incomplete: 0 · skipped checks: 0 · estimated cost (OpenRouter list prices): $0.06 · turn-limit stops: 1 Pass rate = runs whose gate result is Tasks
Skills
Skipped / errored checksNone. Incomplete runsNone. Process checks (treatment arm, not in deltas)
Adversarial tier (reported only, never gates)None. |
Skill Evaluation: rf-core (narrow, treatment, N=3, z-ai/glm-5.3-flash)Model: OpenRouter spend
GateTrigger gaterf-skill-eval trigger reportModel
Description visibilitySessions whose skill listing showed the skill's description (the rest listed it by name only). Own counts the sessions of the skill's own queries, All every session in this result.
Failing queriesNone. rf-skill-eval reportRuns: 48 · pass: 45 · fail: 3 · incomplete: 0 · skipped checks: 0 · estimated cost (OpenRouter list prices): $0.54 · turn-limit stops: 2 Pass rate = runs whose gate result is Tasks
Skills
Skipped / errored checksNone. Incomplete runsNone. Process checks (treatment arm, not in deltas)
Adversarial tier (reported only, never gates)None. |
Summary
Adds
rf-bddto rf-core: writing and reviewing behaviour-driven specifications in native.robotand.resourcefiles (OpenSpec changeadd-rf-bdd-skill).SKILL.md(169 lines) with 11 house rules, Gherkin input, 11 gotchas; 7 references;assets/robocop-bdd.toml; two runnable example projects (resource strategy, World library) on a fake ATM service.trap-verification.md).--selectis rejected" to "selects nothing".rf-test-architect(plus Codex/OpenCode variants) route BDD work to rf-bdd.sut-bdd-libraryfixture (base,steps,review), a BDD grader with 30 tests and golden solutions, five tasks, a trigger set (11 positive, 11 negative, 4 holdout). One-linecli.pypreflight fix: tasks are selected by fixture root (sut-bdd-library/base→sut-bdd-library).## Unreleased, README and rf-core README updated; skill count 10.Not in this PR
Tasks 1.1, 5.5 and 5.6 (live trigger baseline and tuning, CI-model task baselines) follow in a separate PR. A local spot check with Haiku 5.5 loaded rf-bdd in 3 of 3 runs; the late-return suites passed (5/5, 10/10) with no
.featurefiles.Test plan
--check, governance, content rules, validate-skills, plugin references, drift, docs voice, security scanopenspec validate --all --strictAI assistance
Written with Claude Code (Claude Opus 5.5); the maintainer reviewed the change and directed the scope. Reference porting and the eval fixture were done by subagents and checked by the tests above.
🤖 Generated with Claude Code
https://claude.ai/code/session_01FHEmzL3ErAhGQykKWDx2wP