field-impact

You are computing the blast radius for a field change. The user asks "who depends on Account.Custom_Field__c?" and you answer with a structured report grouped by metadata type.

Read Project Config First

source "${CLAUDE_PLUGIN_ROOT}/hooks/lib/config.sh"
source "${CLAUDE_PLUGIN_ROOT}/hooks/lib/sf-cli.sh"
LWC_SRC="$(sf_config_get '.paths.lwcSource' "$ENV")"
APEX_SRC="$(sf_config_get '.paths.apexSource' "$ENV")"

Input

$ARGUMENTS: required.

• <Object>.<Field> — full reference (e.g., Account.Risk_Score__c)
• <Field> — bare field; the skill greps without object qualifier (less precise)
• --target-org <alias> / --env <name> — for org-side checks (formula references, etc.)
• --ci / --format json|sarif / --out <path>

Steps

1. Validate the field exists (if --target-org/cache available)

• Read ${CLAUDE_PLUGIN_DATA}/argo/org-cache/<org>.json if present
• If absent, run sf_cli_describe <Object> <ORG> and extract field list
• If the field doesn't exist, warn but continue with source-only search

2. Search source by metadata type

For each metadata type, define a search and capture file + line + snippet. Use Grep (ripgrep-backed) for speed.

For the bare <Field> form (no object qualifier), prefix matches with the object name to reduce false positives — but warn the user that some references may be missed if the field is referenced as a string.

3. Org-side queries (optional, if --target-org)

# Formula fields on OTHER objects that reference this field
sf_cli_query "
  SELECT EntityDefinition.QualifiedApiName, QualifiedApiName, FormulaDescription
  FROM FieldDefinition
  WHERE FormulaDescription LIKE '%<Object>.<Field>%'
" "$ORG"

# Flow definitions referencing this field (best-effort; relies on flow XML being in source)
# (skipped — covered by source search)

4. Output

Default Markdown:

# Field Impact: Account.Risk_Score__c

Run at: 2026-04-28T11:00:00Z
Field exists in target org: ✅ (custom number, length 18, scale 2)
Total references: 17

## Apex (5)
- `force-app/main/default/classes/AccountController.cls:42` — `acct.Risk_Score__c`
- ...

## LWC JS (3)
- `force-app/main/default/lwc/acmeAcctRisk/acmeAcctRisk.js:12` — `'Account.Risk_Score__c'`
- ...

## LWC HTML (1)
- `force-app/main/default/lwc/acmeAcctRisk/acmeAcctRisk.html:8` — `record.fields.Risk_Score__c.value`

## Page Layouts (4)
- ...

## Validation Rules (1)
- `force-app/main/default/objects/Account/validationRules/Block_Negative_Risk.validationRule-meta.xml`

## Formula Fields (1)
- `Account.Risk_Tier__c` (formula on the same object) — references `Risk_Score__c`

## Flows (2)
- `Account_Risk_Set` — uses Risk_Score__c in a decision element
- ...

## Reports (0)
None in source. (Reports may live only in setup if not source-tracked.)

## Org-only (formula fields on other objects)
- `Opportunity.Account_Risk__c` (formula `Account.Risk_Score__c`)

CI JSON / SARIF:

• Each reference becomes one finding with ruleId: "FIELD-IMPACT", severity: "note", file/line populated, message describes the metadata type
• Exit 0 always (this is informational, not a check) unless --fail-if-found is passed (then 1 if any references)

5. Exit codes

• 0 — informational mode (default), or --fail-if-found and no references found
• 1 — --fail-if-found and references exist
• 2 — query/setup error

Rules

• Be precise about false positives. When grepping by bare field name, the result may include unrelated symbols. Always prefix the match with <Object>. when checking Apex. Note any unresolvable matches in a "Possible matches" section
• Skip generated files. package.xml, deployment manifests, log files
• No string assertEquals on field name. Don't bother flagging test class assertions — they're real references but not interesting for impact analysis. Group them in a "Tests (n)" section without per-line detail
• Time-bound the org-side queries. If the FieldDefinition query times out (large orgs), report the partial result and continue — don't fail the whole run