Skip to content

feat: remove trigger guard + proactive opt-out prompt#457

Merged
garrytan merged 5 commits intomainfrom
garrytan/fix-startup-noise
Mar 25, 2026
Merged

feat: remove trigger guard + proactive opt-out prompt#457
garrytan merged 5 commits intomainfrom
garrytan/fix-startup-noise

Conversation

@garrytan
Copy link
Owner

Summary

  • Removed the MANUAL TRIGGER ONLY prefix from every skill description — was wasting 58 characters per skill and causing Codex build errors
  • Added one-time proactive opt-out prompt: first time a user runs any gstack skill, they're asked if they want proactive skill suggestions (saved as gstack-config set proactive true/false)
  • Fixed telemetry source tagging and duration guards in gstack-telemetry-log

Test Coverage

All new code paths have test coverage. Tests: all pass (exit 0).

Pre-Landing Review

  • Source changes in scripts/gen-skill-docs.ts (6 lines removed), scripts/resolvers/preamble.ts (31 lines added), bin/gstack-telemetry-log (telemetry fix)
  • 26 SKILL.md files regenerated (mechanical — trigger guard removed, proactive prompt added to preamble)
  • No SQL, no LLM trust boundary changes, no security concerns

Test plan

  • All bun tests pass (exit 0)
  • gen-skill-docs regeneration verified
  • VERSION/package.json sync verified

🤖 Generated with Claude Code

garrytan and others added 4 commits March 24, 2026 12:31
Add --source, --error-message, --failed-step flags to gstack-telemetry-log.
Source tagging (live vs test via GSTACK_TELEMETRY_SOURCE env) prevents E2E
tests from polluting production data. Duration guards cap unreasonable
values (>24h or negative → null).

Partial cherry-pick from garrytan/community-mode — non-breaking parts only.
Skips install_fingerprint rename (needs schema migration).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Remove "MANUAL TRIGGER ONLY" injection from all skill descriptions. This
frees 59 chars per skill from the 1024-char Codex description budget and
lets skills auto-fire based on semantic matching.

Merge auto-fire control into the existing `proactive` setting — when false,
Claude won't auto-invoke skills or suggest them. Users are prompted once
about this preference (chains after the telemetry prompt, fires on second
skill run).

Also trims the root gstack description by removing the skill catalog
(already in the body), saving ~500 chars.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@github-actions
Copy link

github-actions bot commented Mar 24, 2026

E2E Evals: ❌ FAIL

82/86 tests passed | $14.07 total cost | 12 parallel runners

Suite Result Status Cost
e2e-browse 7/8 $0.38
e2e-deploy 4/4 $0.57
e2e-design 7/7 $1.89
e2e-plan 11/11 $3.45
e2e-qa-bugs 3/3 $1.32
e2e-qa-workflow 4/4 $1.36
e2e-review 7/7 $1.82
e2e-routing 11/11 $2.12
e2e-workflow 4/7 $0.68
llm-judge 24/24 $0.48

12x ubicloud-standard-2 (Docker: pre-baked toolchain + deps) | wall clock ≈ slowest suite

Failures

  • ❌ browse basic commands: error_max_turns
  • ❌ /ship local workflow: success
  • ❌ /ship local workflow: success
  • ❌ /ship local workflow: success

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@garrytan garrytan merged commit 8500136 into main Mar 25, 2026
18 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant