Salesforce Apex Log Analyzer Skill

salesforce-apex-log-analyzer-skill

T1 read-only runtime skill for Apex debug log retrieval and analysis. This skill is a diagnostic flashlight — it retrieves log content, identifies governor-limit hits, surfaces SOQL N+1 patterns, traces exceptions, and produces a prioritized finding report. It does not execute code, mutate data, or deploy anything.

When This Skill Owns the Task

Use salesforce-apex-log-analyzer-skill when the work requires log-based diagnosis:

• "Why is my Account trigger hitting a governor limit?"
• "Analyze the debug log from my sandbox — it shows a heap size limit"
• "This trigger is slow — can you read the log and find the SOQL bottleneck?"
• "I got an unhandled exception in my batch job — here is the log"
• "Check if my Queueable is completing successfully in the async log"
• "Retrieve the latest log for user jsmith@myorg.sandbox and diagnose it"

Delegate elsewhere when:

Required Context to Gather First

Before retrieving or analyzing any log, confirm:

1. Target org alias — the --target-org value from sf org list.
2. Log identifier — log ID, log file path, user/transaction context, or timeframe. Accept a pasted log file if the user provides one directly.
3. Transaction context — which trigger, class, batch job, or user action generated the log.
4. Goal — governor-limit diagnosis, exception trace, performance analysis, or all.
5. Sensitivity level — does the log likely contain PII (email, phone, record data from regulated objects)? Apply stricter redaction if so.

Recommended Workflow

Step 1 — Verify org alias and reachability

sf org display --target-org <alias>

Confirm the org is reachable. Note org type (production vs sandbox) for the audit envelope. Apply stricter scrutiny for production org logs — log content may contain PII.

Step 2 — List available logs

sf apex get log --list --target-org <alias>

Identify the relevant log(s) by user, timestamp, and log type. If the user provides a log ID, skip this step and proceed to Step 3.

Step 3 — Retrieve the log

Option A — specific log by ID:

sf apex get log \
  --log-id <logId> \
  --target-org <alias>

Option B — latest log for current user:

sf apex get log \
  --number 1 \
  --target-org <alias>

Option C — tail live logs (non-blocking diagnostic):

sf apex tail log \
  --target-org <alias> \
  --color

Option D — user has pasted log content directly: Accept the pasted content and proceed to Step 4.

Step 4 — Parse the log in order

Analyze in this sequence per references/log-format-reference.md:

1. Entry point — identify transaction type (Apex class invocation, trigger, REST callout, batch execute)
2. Fatal errors and unhandled exceptions — FATAL_ERROR and EXCEPTION_THROWN lines
3. Governor limit hits — LIMIT_USAGE_FOR_NS entries; compare to limits in references/governor-limit-signatures.md
4. SOQL patterns — SOQL_EXECUTE_BEGIN / SOQL_EXECUTE_END pairs; look for repeated queries in loop context
5. DML patterns — DML_BEGIN / DML_END pairs; check for loop context
6. CPU hotspots — high cumulative CPU entries; identify top consumers
7. Heap usage — HEAP_ALLOCATE patterns; identify large collection accumulation
8. Async job indicators — ENTERING_MANAGED_PKG, CALLOUT_REQUEST, FUTURE_CALL_PROCESS

Step 5 — Classify findings by severity

Step 6 — Apply redaction

Apply all redaction rules (see Redaction Rules section) before emitting any log excerpt in output. Never echo raw log lines containing record IDs, user IDs, or PII field values.

Step 7 — Emit findings with audit envelope

Score findings against the quality rubric (see below). Emit the full audit envelope. Propose the smallest correct fix for each Critical and Warning finding.

Step 8 — Route to repair skill

If a code fix is needed, hand off to salesforce-apex-generator-skill with the specific finding as context. If a deeper test run is needed, route to salesforce-apex-test-runner-skill.

Quality Scoring Rubric (100-point)

Score the analysis quality before presenting. Threshold: 80+ acceptable.

Scoring penalties:

• Raw log lines with record IDs echoed: -20 (immediate caveat)
• Governor limit hit identified but not traced to source: -15
• SOQL in loop not flagged: -15
• Missing audit envelope: -15
• PII field values in output: score voided (immediate reject)

T1 Least-Privilege Contract

This skill operates at T1 — read-only runtime.

• OAuth scopes: api and refresh_token only.
• Run As account permissions:
  • REQUIRED: View Setup and Configuration
  • DENIED: ModifyAllData, ViewAllData, ViewEncryptedData, ModifyMetadata, AuthorApex, ManageConnectedApps
• No View All Data: Unlike the test runner skill, this skill does NOT require View All Data. Log retrieval uses sf apex get log which operates under the View Setup and Configuration permission scope.
• Read-only: Retrieves log content only. No code execution, no DML, no metadata change.
• Log content sensitivity: Debug logs may contain field values from records processed during the transaction. Apply PII redaction rules strictly.

Audit Envelope Schema

audit_envelope:
  matter_id: "<caller-provided-or-generated-uuid>"
  skill_id: "salesforce-apex-log-analyzer-skill"
  skill_version: "0.1.0"
  target_org_alias: "<alias>"
  run_as_user_id: "<user_id_placeholder>"
  org_type_verified: "sandbox | production | unknown"
  log_id: "<log-id-or-hash-if-pii-risk>"
  log_size_bytes: <integer>
  transaction_entry_point: "<class or trigger name>"
  timestamp: "<ISO-8601-UTC>"
  pii_risk_assessed: true | false
  redactions_applied:
    - field_or_pattern: "<pattern>"
      reason: "<record_id|user_id|pii|session_token>"
  findings_count:
    critical: <integer>
    warning: <integer>
    info: <integer>

Output Format

verdict: "acceptable | caveat | reject"
quality_score: <0-100>
quality_notes: "<scoring rationale>"

transaction_summary:
  entry_point: "<class/trigger/batch>"
  transaction_type: "<trigger|class|batch|queueable|schedulable|rest>"
  org_alias: "<alias>"
  org_type: "sandbox | production | unknown"

findings:
  critical:
    - finding: "<title>"
      location: "<ClassName.method line N>"
      evidence: "<redacted log excerpt>"
      governor_limit: "<limit name if applicable>"
      current_value: <integer>
      limit_ceiling: <integer>
      suggested_fix: "<specific remediation>"
      route_to: "<skill or action>"
  warning:
    - finding: "<title>"
      location: "<ClassName.method line N>"
      evidence: "<redacted log excerpt>"
      suggested_fix: "<specific remediation>"
  info:
    - finding: "<title>"
      notes: "<optimization or hygiene note>"

soql_analysis:
  total_queries: <integer>
  governor_limit: 100
  n_plus_1_patterns:
    - query_excerpt: "<SELECT ... FROM ...>"
      invocation_count: <integer>
      loop_context: "<loop location>"
      suggested_fix: "<move SOQL outside loop; use Map pattern>"
  non_selective_queries: ["<query excerpt>"]

cpu_analysis:
  total_cpu_ms: <integer>
  cpu_limit_ms: 10000
  hotspots: ["<ClassName.method: N ms>"]

heap_analysis:
  peak_heap_bytes: <integer>
  heap_limit_bytes: 6291456
  large_allocations: ["<description>"]

audit_envelope:
  <see Audit Envelope Schema>

next_steps:
  - "<Critical/Warning: route to salesforce-apex-generator-skill for fix>"
  - "<if tests needed after fix: salesforce-apex-test-runner-skill>"

assumptions:
  - "<explicit list>"

missing_evidence:
  - "<what additional log context would help>"

Redaction Rules

Apply in order before emitting any output. Never bypass for any reason.

1. OAuth tokens, refresh tokens, session IDs: Never include. Strip from any CLI output.
2. Salesforce Org IDs (18-char starting with 00D): Replace with <org_id_placeholder>.
3. Salesforce Record IDs (15/18-char alphanumeric): Replace with <record_id_placeholder> in all log excerpts. Record IDs appear frequently in USER_DEBUG, DML_BEGIN, and SOQL_EXECUTE_BEGIN lines.
4. User IDs (OwnerId, CreatedById, Running User ID, User.Id): Replace with <user_id_placeholder>.
5. Email addresses in log output: Replace with <email_placeholder> unless the user explicitly acknowledges PII scope and the org is non-production.
6. Phone numbers, SSNs, financial account numbers in log output: Replace with <pii_placeholder>. Flag in redactions_applied.
7. Session tokens and access tokens in USER_DEBUG lines: Strip entirely. Do not include a placeholder that implies a token value was present.
8. Instance URLs: Replace with org alias in output. Do not emit raw instance URLs.
9. Stack traces: Retain class names and line numbers — required for diagnosis. Replace any record IDs embedded in exception messages.

Handoff Rules

Required handoff fields: matter_id, audit_envelope, findings (sanitized), next_steps.

Stop Conditions

Stop and do not continue if:

• Log retrieval fails and the user cannot provide log content directly — stop and explain that a log ID, user context, or pasted log content is required.
• Log content contains fields identified as encrypted (Shield PE / PMLE) — skip those entries, note the redaction, and analyze the remainder.
• The audit envelope cannot be completed — stop until matter_id or org alias is resolved.
• The user requests redaction to be disabled — stop and explain the policy.
• Log content is from a production org and contains PII fields — apply maximum redaction; if PII is pervasive and the matter classification does not permit, stop and escalate to the compliance specialist.

Security Notes

• T1 read-only runtime: No code execution, no DML, no metadata mutation.
• No View All Data required: Log retrieval operates under View Setup and Configuration only. This distinguishes this skill from the test runner and reduces its permission footprint.
• Log content sensitivity: Apex debug logs may capture field values, user data, and API payloads. PII and encrypted field redaction is mandatory.
• Sanitized output only: All record IDs, user IDs, session tokens, and PII values redacted before emission. Raw log lines are never echoed.
• Structured audit: Every execution produces a complete audit envelope including pii_risk_assessed and redactions_applied fields.
• Revocable: Rotating the Run As account's refresh token immediately revokes all access.

Reference File Index