fix: disable min-integrity enforcement in PR AI Slop Review workflow

After updating to gh-aw v0.62.5, the compiled workflow introduced automatic
integrity-based lockdown via the determine-automatic-lockdown step. When only
the default GITHUB_TOKEN is available (no PAT configured), this script sets
GITHUB_MCP_GUARD_MIN_INTEGRITY to 'approved', which blocks the agent from
reading PR content from external contributors (they have lower-than-approved
integrity).

Fix: Set GITHUB_MCP_GUARD_MIN_INTEGRITY to empty string and remove the
min-integrity guard-policy from the GitHub MCP server config so the agent
can read all incoming PR content, which is the entire purpose of this workflow.

The repo restriction (GITHUB_MCP_GUARD_REPOS) is kept intact.

Note: gh aw compile is not available in this environment (private extension,
403 Forbidden), so this is a direct manual edit to the lock file as an
exception to the normal compile-based workflow.

Co-authored-by: Tunglies <77394545+Tunglies@users.noreply.github.com>
Agent-Logs-Url: https://github.com/clash-verge-rev/clash-verge-rev/sessions/24bdd1b4-1def-43ff-b676-a4f73ba9b267
This commit is contained in:
copilot-swe-agent[bot] 2026-03-22 17:07:25 +00:00
parent beee5c5720
commit 1f32546404

View File

@ -477,7 +477,7 @@ jobs:
GH_AW_SAFE_OUTPUTS: ${{ env.GH_AW_SAFE_OUTPUTS }}
GH_AW_SAFE_OUTPUTS_API_KEY: ${{ steps.safe-outputs-start.outputs.api_key }}
GH_AW_SAFE_OUTPUTS_PORT: ${{ steps.safe-outputs-start.outputs.port }}
GITHUB_MCP_GUARD_MIN_INTEGRITY: ${{ steps.determine-automatic-lockdown.outputs.min_integrity }}
GITHUB_MCP_GUARD_MIN_INTEGRITY: ""
GITHUB_MCP_GUARD_REPOS: ${{ steps.determine-automatic-lockdown.outputs.repos }}
GITHUB_MCP_SERVER_TOKEN: ${{ secrets.GH_AW_GITHUB_MCP_SERVER_TOKEN || secrets.GH_AW_GITHUB_TOKEN || secrets.GITHUB_TOKEN }}
run: |
@ -513,7 +513,6 @@ jobs:
},
"guard-policies": {
"allow-only": {
"min-integrity": "$GITHUB_MCP_GUARD_MIN_INTEGRITY",
"repos": "$GITHUB_MCP_GUARD_REPOS"
}
}