mirror of
https://github.com/forcedotcom/afv-library.git
synced 2026-07-30 03:09:50 +08:00
W-21701715: Enhance Expert Global Rule (#86)
* refactor: enhance expert global rule * refactor: enhance expert global rule
This commit is contained in:
Hemant Singh Bisht
GitHub
parent
5c40e3b63e
commit
cf7f55c3f8
@ -1,73 +1,88 @@
|
||||
# Rule: Salesforce Metadata Generation
|
||||
|
||||
## Objective
|
||||
Enforce: **skill load → API context → file generation** for all Salesforce metadata.
|
||||
Enforce: **skill load -> API context -> file generation** for all Salesforce metadata.
|
||||
|
||||
## Constraints
|
||||
|
||||
1. **Never write** without both: metadata type skill loaded AND `Salesforce API Context` MCP Server called for that type
|
||||
2. **One type at a time** - complete full cycle before next type
|
||||
3. **Child types need own context** - if adding any child metadata inside a parent metadata's file, load skill and call `Salesforce API Context` for each child type (e.g. CustomField inside CustomObject) separately; don't rely on the parent's schema for creating child metadata
|
||||
4. **Max one clarifying question** before starting
|
||||
5. **Don't call `execute_metadata_action` unless a skill instructs to do so**
|
||||
1. **Never write** without a loaded metadata type skill for that type.
|
||||
2. **One type at a time** - complete the full cycle for the current type before moving to the next type.
|
||||
3. **Always attempt `salesforce-api-context` MCP** for each type before writing; if it is unavailable after a real attempt, proceed with the loaded skill only.
|
||||
4. **Child types need their own API context response** - if adding child metadata inside a parent metadata file, load the child metadata skill and use `salesforce-api-context` MCP for each child type separately; do not rely on the parent's schema or API context response for child metadata creation. The same fallback in constraint 3 applies.
|
||||
5. **Max one clarifying question** before starting.
|
||||
6. **Do not call `execute_metadata_action` unless a skill instructs you to do so.**
|
||||
|
||||
## Workflow
|
||||
## Skill Selection
|
||||
|
||||
### 1. Detect Intent
|
||||
- **App skill** (end-to-end capability like lex/react App) → App Path
|
||||
- **Direct metadata** (specific fields/objects/pages) → Direct Path
|
||||
- Search available skills and load the best-matching app-level skill first when the request asks for a Lightning app, end-to-end solution, or a business app spanning multiple metadata types.
|
||||
- Search available skills and load the best-matching per-type metadata skill first when the request targets individual metadata components.
|
||||
|
||||
### 2. App Path
|
||||
1. Load App related skill, extract metadata type sequence
|
||||
2. For each type, execute loop (a-e below)
|
||||
3. Proceed to Step 3
|
||||
## Initial Gate
|
||||
|
||||
### 2. Direct Path
|
||||
1. Identify all needed types
|
||||
2. For each type in dependency order, execute loop (a-e below)
|
||||
3. Proceed to Step 3
|
||||
Never create files or generate metadata before completing skill selection.
|
||||
|
||||
### Loop (a-e) - Execute for Each Type
|
||||
1. Determine whether the request is app-level or metadata-type-level.
|
||||
2. Search available skills for the request and identify the best-matching candidate skill.
|
||||
3. If the request is app-level, load the best-matching app-level skill by reading its contents and use it to identify metadata types and dependency order.
|
||||
4. If the request is metadata-type-level, identify the target metadata type and the best-matching per-type metadata skill for that type. Do not treat skill selection as the per-type skill-load step.
|
||||
5. Confirm skill selection with:
|
||||
`intent=<app|type> selected_skill=<exact-skill-name|none> skill_selection=complete|pending`
|
||||
6. Set `skill_selection=complete` only after the exact selected skill name has been identified and recorded.
|
||||
7. Print this exact skill-selection status line in the chat before proceeding.
|
||||
|
||||
Do not continue until `skill_selection=complete` and `selected_skill=<exact-skill-name|none>` are recorded.
|
||||
|
||||
## Per-Type Loop (a-e)
|
||||
|
||||
For each metadata type in scope, whether identified by an app-level skill or requested directly, execute steps a through e below one metadata type at a time. Do not create or modify files for the current metadata type, and do not move to the next metadata type, until steps a through e are complete.
|
||||
|
||||
**a. Load Skill**
|
||||
- Load metadata type skill once (not per record)
|
||||
- If no skill exists, continue with API context only
|
||||
- Search available skills for the best-matching skill for the current metadata type and load it by reading its SKILL.md.
|
||||
- Record `selected_skill=<exact-skill-name|none>` for the current metadata type before proceeding.
|
||||
- Load once per type, not per record.
|
||||
- If no skill exists after a real search, stop and ask for guidance instead of writing without a skill.
|
||||
|
||||
**b. Call API Context**
|
||||
- Use `Salesforce API Context` and make use of these tools as per requirement:
|
||||
- `get_metadata_type_sections`
|
||||
- `get_metadata_type_context`
|
||||
- `get_metadata_type_fields`
|
||||
- `get_metadata_type_fields_properties`
|
||||
- `search_metadata_types`
|
||||
**b. Use `salesforce-api-context` MCP**
|
||||
- Use one or more of these tools as required:
|
||||
- `get_metadata_type_sections`
|
||||
- `get_metadata_type_context`
|
||||
- `get_metadata_type_fields`
|
||||
- `get_metadata_type_fields_properties`
|
||||
- `search_metadata_types`
|
||||
- Attempt API context for every type before writing.
|
||||
- Record which tools were called for the current metadata type before proceeding.
|
||||
- If API context is unavailable after a real attempt, record that state and continue with the loaded skill only.
|
||||
- Treat templates, generators, pipelines, and CLI bootstraps as implementation aids, not as replacements for this gate.
|
||||
|
||||
**c. Pre-Write Gate**
|
||||
- Before EVERY write: confirm API context called for this type
|
||||
- If no → stop and call now
|
||||
- Before EVERY write: confirm `selected_skill=<exact-skill-name>` is recorded and that skill is loaded for this type.
|
||||
- Before EVERY write: confirm `mcp_tools=<tool-list>` is recorded for this type, or confirm `mcp=unavailable` after a real attempt.
|
||||
- If `selected_skill=none` or the skill is missing -> stop.
|
||||
- If `mcp_tools=none` and `mcp` is not `unavailable` -> stop and call MCP now.
|
||||
|
||||
**d. Generate Files**
|
||||
- Use skill constraints + API context
|
||||
- Generate all records for this type now
|
||||
- Use skill constraints + API context when available.
|
||||
- Use skill constraints only when API context was unavailable after a real attempt.
|
||||
- Generate all records for this type now.
|
||||
|
||||
**e. Checkpoint**
|
||||
- Skill loaded? API context called? All files written?
|
||||
- Only proceed to next type when all true
|
||||
|
||||
### 3. Deploy Verification
|
||||
```bash
|
||||
sf project deploy start --dry-run -d "force-app/main/default" --target-org <alias> --test-level NoTestRun --wait 10 --json
|
||||
```
|
||||
On failure: attempt to fix the errors and re-run, retrying up to a maximum of 3 times until it succeeds.
|
||||
- Record:
|
||||
`type=<metadata-type> selected_skill=<exact-skill-name|none> skill=complete|pending mcp=complete|pending|unavailable mcp_tools=<tool-list|none>`
|
||||
- Skill loaded? API context called or unavailable after a real attempt? All files written?
|
||||
- Only proceed to the next type when all are true.
|
||||
|
||||
## Anti-Patterns
|
||||
|
||||
| Don't | Why | Do |
|
||||
|-------|-----|-----|
|
||||
| Write without API context | Missing schema validation | Call API context before first write |
|
||||
| Reload skill per record | Wastes tokens | Load once per type |
|
||||
| Skip API context for later types | No schema for those types | Call for EVERY type |
|
||||
| Skip metadata skills | Missing platform constraints | Load skill for every type |
|
||||
| Ask 3+ questions | Token waste | Max 1 question |
|
||||
| Skip App skill gates | Wrong artifacts | Follow all mandatory gates |
|
||||
| Write despite missing checkpoint | Aware violation | Stop and complete missing step |
|
||||
| Batch types in API call | Violates constraint #3 | One type per call |
|
||||
| Never write without loading the metadata skill | Missing platform constraints | Search for and load the skill before any write |
|
||||
| Never mark `skill_selection=complete` without `selected_skill=<exact-skill-name|none>` | Fake gate completion | Record the exact selected skill before continuing |
|
||||
| Never treat skill selection as skill loading | Fake gate completion | Perform the actual per-type skill load in step a |
|
||||
| Never skip the Initial Gate | Sequence breach | Complete skill selection before any generation |
|
||||
| Never reload a skill per record | Wastes tokens | Load once per type |
|
||||
| Never skip the API context attempt for any type | No schema for those types | Attempt API context for EVERY type |
|
||||
| Never write using API context alone without a loaded skill | Missing platform constraints | Stop and ask for guidance when no skill exists |
|
||||
| Never write without recorded `mcp_tools` or `mcp=unavailable` | No evidence of MCP gate completion | Record MCP tool usage before any write |
|
||||
| Never ask more than 1 clarifying question | Token waste | Max 1 question |
|
||||
| Never skip any gate in the loop (skill load, API context, pre-write, checkpoint) | Wrong artifacts | Follow all mandatory gates in the loop (a-e) |
|
||||
| Never write with a missing checkpoint | Aware violation | Stop and complete missing step |
|
||||
|
||||
Loading…
Reference in New Issue
Block a user