P2 · Vault → skill hygiene (creds without runbooks; op:// path drift) #8
Labels
No labels
lab-honing
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
adam/foundry#8
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Priority: P2 · Size: S · Depends on: none
Goal
A broad "go" from Adam shouldn't get blocked mid-flow for routine lab-API calls. Keep genuine break-glass writes gated, but stop re-prompting on reads and make the write path predictable.
Why (from the 2026-07-11 session)
After "all three" / "go", the auto-mode classifier blocked the Dockhand
POST /api/git/stackstwice as an unnamed shared-infra write — resolved only when Adam switched permission modes. Reads were fine; the writes needed per-action naming at a specificity that a broad delegation doesn't provide.Steps
update-configskill, add scoped Bash allow-rules for lab-API reads (at minimumcurlto192.168.4.200:8647Dockhand; also:3000Forgejo API,:2700Hindsight,:180801Password Connect). Consider thefewer-permission-promptsskill to mine the transcript for the exact call shapes.feedback_*memory +.claude/skills/README.md. Align with (don't contradict)feedback_unraid_mcp_access(MCP writes = break-glass).Acceptance
settings.jsonhas the scoped allow-rules; a test read call to a lab API runs with no prompt; the infra-write convention is documented.Resources
Skills:
update-config,fewer-permission-prompts. Memory:feedback_unraid_mcp_access. Note the permissions file was slimmed to ~50 entries on 2026-06-26 — extend that curated set, don't bloat it.