Conversation
Agent-Logs-Url: https://github.com/github/gh-aw/sessions/8b811642-c8da-42c3-a77e-787c24bebda3 Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
Agent-Logs-Url: https://github.com/github/gh-aw/sessions/8b811642-c8da-42c3-a77e-787c24bebda3 Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
Agent-Logs-Url: https://github.com/github/gh-aw/sessions/8b811642-c8da-42c3-a77e-787c24bebda3 Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
There was a problem hiding this comment.
Pull request overview
Adds a new daily Agentic Workflows pipeline that runs fastxyz/skill-optimizer in a dedicated pre-agent job, hands results to the agent via an artifact, and has the agent create a single issue containing exactly three improvement recommendations.
Changes:
- Introduces a new
daily-skill-optimizer.mdworkflow with askill_optimizerjob that clones/builds/runsfastxyz/skill-optimizerand uploads logs +summary.jsonas an artifact. - Configures the agent phase to download the artifact and enforce a “one issue / exactly three improvements / evidence from artifacts” contract via Safe Outputs.
- Adds the compiled workflow output (
daily-skill-optimizer.lock.yml) aligned to the markdown workflow definition.
Show a summary per file
| File | Description |
|---|---|
| .github/workflows/daily-skill-optimizer.md | Defines the pre-agent skill_optimizer job, artifact handoff, and agent prompt + safe-outputs constraints for issue creation. |
| .github/workflows/daily-skill-optimizer.lock.yml | Generated compiled workflow reflecting the new markdown workflow topology and configuration. |
Copilot's findings
Tip
Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
- Files reviewed: 2/2 changed files
- Comments generated: 3
| env: | ||
| OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }} | ||
| run: | |
There was a problem hiding this comment.
OPENROUTER_API_KEY is set at the step level, so it will also be visible to npm ci/npm run build and other third-party code executed in this step (including npm lifecycle scripts). To reduce secret exposure, consider scoping this secret to only the benchmark invocation (e.g., export it immediately before the non---dry-run run, or move the benchmark run into a separate step with env:).
| RESULT_DIR="/tmp/gh-aw/skill-optimizer-results" | ||
| TOOL_DIR="$RESULT_DIR/skill-optimizer-src" | ||
| mkdir -p "$RESULT_DIR" | ||
|
|
||
| git clone --depth 1 https://github.com/fastxyz/skill-optimizer "$TOOL_DIR" >"$RESULT_DIR/clone.log" 2>&1 |
There was a problem hiding this comment.
The agent phase requires summary.json, but with set -euo pipefail a failure in git clone/npm ci/npm run build will exit the step before summary.json is written. Consider ensuring summary.json is always produced (e.g., via a trap/finally block that writes a failure summary including which phase failed), and capture exit codes for clone/build similarly to init/run.
| uses: actions/upload-artifact@v7.0.1 | ||
| with: | ||
| name: skill-optimizer-results | ||
| path: /tmp/gh-aw/skill-optimizer-results |
There was a problem hiding this comment.
The artifact upload path points at /tmp/gh-aw/skill-optimizer-results, which also contains the cloned skill-optimizer-src checkout (and its node_modules after npm ci). This can greatly inflate artifact size and slow/fail uploads. Consider keeping the tool checkout outside the artifact directory or deleting skill-optimizer-src before uploading, and/or uploading only the specific result files needed by the agent (e.g., summary.json + logs).
| path: /tmp/gh-aw/skill-optimizer-results | |
| path: | | |
| /tmp/gh-aw/skill-optimizer-results/summary.json | |
| /tmp/gh-aw/skill-optimizer-results/clone.log | |
| /tmp/gh-aw/skill-optimizer-results/npm-ci.log | |
| /tmp/gh-aw/skill-optimizer-results/npm-build.log | |
| /tmp/gh-aw/skill-optimizer-results/init.log | |
| /tmp/gh-aw/skill-optimizer-results/run.log | |
| /tmp/gh-aw/skill-optimizer-results/benchmark.log |
This introduces a daily agentic workflow that runs
fastxyz/skill-optimizerin a dedicated pre-agent job, then uses its artifact output to drive one actionable issue with exactly three improvement recommendations.Workflow topology (separate execution + handoff)
daily-skill-optimizer.mdwith a dedicatedskill_optimizerjob (needs: [activation]).fastxyz/skill-optimizer, runs initialization + benchmark/dry-run flow, and writes structured results (summary.json) plus logs.Agent behavior (issue creation contract)
safe-outputs.create-issuewith a single-issue cap.summary.json.Compiled workflow output
daily-skill-optimizer.lock.yml) aligned to the new markdown workflow definition.🤖 Smoke CI scheduled run completed — https://github.com/github/gh-aw/actions/runs/24814493092