beads: remove pending_approval/pending_review workflow
Bots no longer use discrete pending states for human gates. Substantive reviews become child beads; quick checkpoints are description-driven. Remove mark_pending_approval.sh, mark_pending_review.sh, resume_work.sh and purge the approval-gate section from BEADS.md. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Levi Neely
parent
4de14a2159
commit
c57e98e167
69
BEADS.md
69
BEADS.md
|
|
@ -27,7 +27,7 @@ anvillm/beads/
|
|||
├── label/<label> # Issues with label (JSON)
|
||||
├── children/<id> # Direct children of parent (JSON)
|
||||
└── <bead-id>/
|
||||
├── status # open | in_progress | pending_approval | pending_review | closed
|
||||
├── status # open | in_progress | closed
|
||||
├── title # Bead title
|
||||
├── description # Bead description
|
||||
├── assignee # Assigned actor
|
||||
|
|
@ -62,9 +62,6 @@ anvillm/beads/
|
|||
| `label` | `label <bead-id> 'label'` | Add label to bead |
|
||||
| `unlabel` | `unlabel <bead-id> 'label'` | Remove label from bead |
|
||||
| `set-capability` | `set-capability <bead-id> low\|standard\|high` | Set capability level (replaces existing) |
|
||||
| `pending-approval` | `pending-approval <bead-id> [assignee]` | Set status=pending_approval and assignee (default: user) |
|
||||
| `pending-review` | `pending-review <bead-id> [assignee]` | Set status=pending_review and assignee (default: user) |
|
||||
| `resume` | `resume <bead-id> [assignee]` | Set status=in_progress and assignee after approval (default: user) |
|
||||
| `init` | `init [prefix]` | Initialize beads with custom ID prefix (default: bd) |
|
||||
| `batch-create` | `batch-create <json-array>` | Create multiple beads from JSON array |
|
||||
|
||||
|
|
@ -188,70 +185,6 @@ Available filter fields:
|
|||
- `labels` (array): Filter by labels (all must match)
|
||||
- `parent_id` (string): Filter by parent ID
|
||||
|
||||
## Approval Gates (Human in the Loop)
|
||||
|
||||
Approval gates allow bots to pause work and request human sign-off before proceeding with critical or irreversible operations.
|
||||
|
||||
### Status Values
|
||||
|
||||
Two statuses mark beads awaiting a human response:
|
||||
|
||||
| Status | Description |
|
||||
|--------|-------------|
|
||||
| `pending_approval` | Bot sent `APPROVAL_REQUEST`; waiting for human `APPROVAL_RESPONSE` |
|
||||
| `pending_review` | Bot sent `REVIEW_REQUEST`; waiting for human `REVIEW_RESPONSE` |
|
||||
|
||||
Beads in either status are **not** surfaced by `anvillm/beads/ready` — bots will not accidentally claim them. Use `anvillm/beads/pending` to list all beads awaiting human input.
|
||||
|
||||
### Label Conventions
|
||||
|
||||
Add these labels to beads that should trigger a human gate at completion:
|
||||
|
||||
| Label | Meaning |
|
||||
|-------|---------|
|
||||
| `requires_approval` | Completion requires an `APPROVAL_REQUEST`/`APPROVAL_RESPONSE` exchange |
|
||||
| `requires_review` | Completion requires a `REVIEW_REQUEST`/`REVIEW_RESPONSE` exchange |
|
||||
|
||||
Example: `echo "label bd-abc requires_approval" | 9p write anvillm/beads/ctl`
|
||||
|
||||
### Approval Workflow
|
||||
|
||||
```sh
|
||||
# 1. Bot sends approval request to user
|
||||
# (via anvillm-communication skill or send_message MCP tool)
|
||||
# type: APPROVAL_REQUEST
|
||||
# subject: "Approve: delete production database backup?"
|
||||
# body: "I am about to run: DROP TABLE backups. Reason: cleanup task bd-xyz. Approve?"
|
||||
|
||||
# 2. Bot atomically marks bead pending and assigns to user for review
|
||||
echo "pending-approval bd-xyz" | 9p write anvillm/beads/ctl
|
||||
# (optionally assign to a specific agent: "pending-approval bd-xyz agent-id")
|
||||
|
||||
# 3. Human reviews in their inbox (Assist, anvilweb, TUI, or Emacs)
|
||||
# and clicks Approve or Reject
|
||||
|
||||
# 4a. On APPROVAL_RESPONSE (approved) — bot resumes and reassigns to itself:
|
||||
echo "resume bd-xyz $AGENT_ID" | 9p write anvillm/beads/ctl
|
||||
# ... continue work ...
|
||||
echo "complete bd-xyz" | 9p write anvillm/beads/ctl
|
||||
|
||||
# 4b. On APPROVAL_RESPONSE (rejected) — bot stops:
|
||||
echo "fail bd-xyz 'human rejected: too risky'" | 9p write anvillm/beads/ctl
|
||||
```
|
||||
|
||||
### Monitoring Pending Approvals
|
||||
|
||||
```sh
|
||||
# List all beads awaiting human input
|
||||
9p read anvillm/beads/pending | jq .
|
||||
|
||||
# Filter only approval-pending
|
||||
9p read anvillm/beads/pending | jq '[.[] | select(.status == "pending_approval")]'
|
||||
|
||||
# Filter only review-pending
|
||||
9p read anvillm/beads/pending | jq '[.[] | select(.status == "pending_review")]'
|
||||
```
|
||||
|
||||
## Initialization
|
||||
|
||||
Initialize beads in a project (creates `.beads/` directory with Dolt database):
|
||||
|
|
|
|||
|
|
@ -1,31 +0,0 @@
|
|||
#!/bin/bash
|
||||
# capabilities: beads
|
||||
# description: Mark a bead as pending approval
|
||||
# Usage: mark_pending_approval.sh --mount <mount> --id <bead-id> [--assignee <id>]
|
||||
set -euo pipefail
|
||||
|
||||
|
||||
MOUNT=""
|
||||
BEAD_ID=""
|
||||
ASSIGNEE=""
|
||||
|
||||
while [[ $# -gt 0 ]]; do
|
||||
case "$1" in
|
||||
--mount) MOUNT="$2"; shift 2 ;;
|
||||
--id) BEAD_ID="$2"; shift 2 ;;
|
||||
--assignee) ASSIGNEE="$2"; shift 2 ;;
|
||||
*) echo "unknown argument: $1" >&2; exit 1 ;;
|
||||
esac
|
||||
done
|
||||
|
||||
if [ -z "$MOUNT" ] || [ -z "$BEAD_ID" ]; then
|
||||
echo "usage: mark_pending_approval.sh --mount <mount> --id <bead-id> [--assignee <id>]" >&2
|
||||
exit 1
|
||||
fi
|
||||
|
||||
if [ -n "$ASSIGNEE" ]; then
|
||||
echo "pending-approval $BEAD_ID $ASSIGNEE" | 9p write anvillm/beads/$MOUNT/ctl
|
||||
else
|
||||
echo "pending-approval $BEAD_ID" | 9p write anvillm/beads/$MOUNT/ctl
|
||||
fi
|
||||
echo "pending-approval: $BEAD_ID${ASSIGNEE:+ → $ASSIGNEE}"
|
||||
|
|
@ -1,31 +0,0 @@
|
|||
#!/bin/bash
|
||||
# capabilities: beads
|
||||
# description: Mark a bead as pending review
|
||||
# Usage: mark_pending_review.sh --mount <mount> --id <bead-id> [--assignee <id>]
|
||||
set -euo pipefail
|
||||
|
||||
|
||||
MOUNT=""
|
||||
BEAD_ID=""
|
||||
ASSIGNEE=""
|
||||
|
||||
while [[ $# -gt 0 ]]; do
|
||||
case "$1" in
|
||||
--mount) MOUNT="$2"; shift 2 ;;
|
||||
--id) BEAD_ID="$2"; shift 2 ;;
|
||||
--assignee) ASSIGNEE="$2"; shift 2 ;;
|
||||
*) echo "unknown argument: $1" >&2; exit 1 ;;
|
||||
esac
|
||||
done
|
||||
|
||||
if [ -z "$MOUNT" ] || [ -z "$BEAD_ID" ]; then
|
||||
echo "usage: mark_pending_review.sh --mount <mount> --id <bead-id> [--assignee <id>]" >&2
|
||||
exit 1
|
||||
fi
|
||||
|
||||
if [ -n "$ASSIGNEE" ]; then
|
||||
echo "pending-review $BEAD_ID $ASSIGNEE" | 9p write anvillm/beads/$MOUNT/ctl
|
||||
else
|
||||
echo "pending-review $BEAD_ID" | 9p write anvillm/beads/$MOUNT/ctl
|
||||
fi
|
||||
echo "pending-review: $BEAD_ID${ASSIGNEE:+ → $ASSIGNEE}"
|
||||
|
|
@ -1,31 +0,0 @@
|
|||
#!/bin/bash
|
||||
# capabilities: beads
|
||||
# description: Resume work on a bead after approval/review
|
||||
# Usage: resume_work.sh --mount <mount> --id <bead-id> [--assignee <id>]
|
||||
set -euo pipefail
|
||||
|
||||
|
||||
MOUNT=""
|
||||
BEAD_ID=""
|
||||
ASSIGNEE=""
|
||||
|
||||
while [[ $# -gt 0 ]]; do
|
||||
case "$1" in
|
||||
--mount) MOUNT="$2"; shift 2 ;;
|
||||
--id) BEAD_ID="$2"; shift 2 ;;
|
||||
--assignee) ASSIGNEE="$2"; shift 2 ;;
|
||||
*) echo "unknown argument: $1" >&2; exit 1 ;;
|
||||
esac
|
||||
done
|
||||
|
||||
if [ -z "$MOUNT" ] || [ -z "$BEAD_ID" ]; then
|
||||
echo "usage: resume_work.sh --mount <mount> --id <bead-id> [--assignee <id>]" >&2
|
||||
exit 1
|
||||
fi
|
||||
|
||||
if [ -n "$ASSIGNEE" ]; then
|
||||
echo "resume $BEAD_ID $ASSIGNEE" | 9p write anvillm/beads/$MOUNT/ctl
|
||||
else
|
||||
echo "resume $BEAD_ID" | 9p write anvillm/beads/$MOUNT/ctl
|
||||
fi
|
||||
echo "resumed $BEAD_ID${ASSIGNEE:+ → $ASSIGNEE}"
|
||||
Reference in New Issue