🔬 Sub-Agent Model Resolution Audit — 2026-10-08
Period: last 2 days · Runs analyzed: 20 · With sub-agents: 1
⚠️ Mismatches Found (0 total)
No size mismatches were observed in the sampled runs.
Unresolved sub-agent dispatch
One run ([§37732970288](https://github.com/github/gh-aw/actions/runs/37732970288)) showed an unresolved dispatch for nano-whoami:
- Declared alias:
gpt-5-nano → expected small
- Observed routing evidence:
fatal: Pi sub-agent "nano-whoami" model alias "gpt-5-nano" did not resolve on github-copilot
- Classification: ❓ unknown (no actual model was selected, so no size comparison was possible)
The same run resolved the other two sub-agents correctly:
haiku-whoami → github-copilot/claude-haiku-4.5 ✅ correct
mini-whoami → github-copilot/gpt-5.4-mini ✅ correct
✅ Correct Resolutions (2 sub-agents verified)
| Workflow |
Sub-agent |
Declared |
Actual Model |
| Smoke Pi Sub Agents |
haiku-whoami |
claude-haiku-4.5 |
github-copilot/claude-haiku-4.5 |
| Smoke Pi Sub Agents |
mini-whoami |
gpt-5-mini |
github-copilot/gpt-5.4-mini |
📋 Analysis Notes
- Downloaded sample contained 20 runs across two log batches.
- Only one sampled run used inline sub-agents.
agent-stdio.log was not directly readable from the cached /tmp/gh-aw/aw-mcp/logs tree due filesystem permissions, so workflow log evidence from gh run view --log was used instead.
- Confidence: high for the observed
haiku-whoami and mini-whoami model resolutions; medium for the unresolved nano-whoami dispatch because no actual model was selected.
References:
Generated by 🔬 Daily Sub-Agent Model Resolution Audit · copilot · gpt54mini · 22.5 AIC · ⌖ 2.52 AIC · ⊞ 14.7K · ◷
🔬 Sub-Agent Model Resolution Audit — 2026-10-08
Period: last 2 days · Runs analyzed: 20 · With sub-agents: 1
No size mismatches were observed in the sampled runs.
Unresolved sub-agent dispatch
One run (
[§37732970288](https://github.com/github/gh-aw/actions/runs/37732970288)) showed an unresolved dispatch fornano-whoami:gpt-5-nano→ expected smallfatal: Pi sub-agent "nano-whoami" model alias "gpt-5-nano" did not resolve on github-copilotThe same run resolved the other two sub-agents correctly:
haiku-whoami→github-copilot/claude-haiku-4.5✅ correctmini-whoami→github-copilot/gpt-5.4-mini✅ correct✅ Correct Resolutions (2 sub-agents verified)
claude-haiku-4.5github-copilot/claude-haiku-4.5gpt-5-minigithub-copilot/gpt-5.4-mini📋 Analysis Notes
agent-stdio.logwas not directly readable from the cached/tmp/gh-aw/aw-mcp/logstree due filesystem permissions, so workflow log evidence fromgh run view --logwas used instead.haiku-whoamiandmini-whoamimodel resolutions; medium for the unresolvednano-whoamidispatch because no actual model was selected.References: