| name | looply-triage |
| description | Use to classify bugs and issues by severity using looply artifacts as primary source. Recommends whether the problem requires formal investigation or immediate action. |
Use this skill when the user explicitly invokes $looply-triage, asks to run /looply:triage, or when triaging bugs/issues before formal investigation.
Workflow phase: diagnosis.
Primary orchestrator: problem-investigator.
Quick usage:
$looply-triage <feature-name> <scope-reference> "[problem-description]" "[constraints...]"
Primary references:
- Workflow playbook: ../../../../.looply/state/workflow-playbook.opencode.md
- Host status contract: ../../../../.looply/state/host-status-contract.json
- Managed pack: ../../../../.looply/managed/packs/software-delivery-suite
- Custom overrides: ../../../../.looply/custom
- Execution hints: ../../../../.looply/state/execution-hints.opencode.json
- Context index: ../../../../.looply/state/context-index.md
- Project context: ../../../../.looply/custom/project-context.md
- Session context: ../../../../.looply/custom/session-context.md
- Code context: ../../../../.looply/state/code-context.json
- Knowledge graph: ../../../../.looply/state/knowledge-graph.json
Usage:
- Explicit mention:
$looply-triage
- Workflow alias to honor:
/looply:triage
- Syntax in Codex:
$looply-triage <feature-name> <scope-reference> "[problem-description]" "[constraints...]"
Example:
- $looply-triage pix-webhook-retry payments-api "webhooks fail intermittently during peak hours" "verify race condition"
When to use:
- User reports a bug without clear severity context
- Before deciding between formal investigation (problem-evaluator) or direct action
- To classify multiple bugs in backlog and prioritize
Curated example guidance:
- ICL mode:
on
- Use examples only for style, structure and quality calibration.
- Do not copy feature-specific names, identifiers or business details from examples.
- No example was selected for this workflow.
Triage Flow
Stage 1: Artifact Triage
- Read relevant looply artifacts:
- feature's workflow-status.md (if it exists)
- associated stories and specs
- code-context.json for affected modules
- knowledge-graph.json for dependencies
- If artifacts are insufficient, inspect the codebase directly.
Stage 2: Severity Classification
Classify using the P0-P3 scale:
P0 - Critical: system unavailable, data loss, security failure
P1 - High: core functionality broken, no workaround
P2 - Medium: functionality partially affected, workaround available
P3 - Low: cosmetic, low impact, rare edge case
Classification criteria:
- End-user impact
- Core vs peripheral functionality
- Workaround availability
- Occurrence frequency
- Data or security at risk
Stage 3: Recommendation
Based on severity:
| Severity | Action |
|---|
| P0 | Escalate immediately to problem-evaluator |
| P1 | Escalate to problem-evaluator with full context |
| P2 | Recommend problem-evaluator OR direct action (if cause is obvious) |
| P3 | Recommend direct action or backlog |
Stage 4: Triage Decision
Generate report in .looply/custom/features/<feature-name>/triage-report.md:
# Triage Report
## Classification
Severity: P1
## Analysis
- Artifacts consulted: stories, code-context.json
- Affected modules: payments-api, webhook-dispatcher
- Probable cause: race condition in consumer
## Recommendation
Escalate to problem-evaluator focusing on race condition in webhook consumer
Execution rules:
- Start by reading the workflow playbook and the feature state file if it exists.
- Run
looply refresh-code-context --check to ensure code-context is fresh.
- If the user asked for help, explain syntax, arguments, example, expected output and next step without mutating state.
- Classify severity using looply artifacts as primary source.
- If cause is obvious from artifacts and severity is P2 or P3, recommend direct action instead of escalating.
- Use managed pack files as canonical process definition and write local state only under
.looply/custom.
- Respond in the same language as the user's input.
- For existing projects, use the real local codebase as the primary source of truth and use context files only as accelerators.
- If a context file has
status: empty, status: draft or status: stale, validate it against the local codebase before trusting it.
- Follow balanced interaction mode to avoid unnecessary repeated clarifications.
- When curated examples are referenced, use them only for style, structure and quality calibration.
- Keep the response visually structured with clear Markdown section titles.
- Do not use emojis.
Composed Agent Context
The sections below were pre-composed by looply from agent context_slots.
Constraints
- Do not implement fixes during triage
- Do not invent causes without evidence from artifacts or codebase
- Use looply artifacts as primary source before deep diving into codebase
Escalation
- Escalate P0/P1 to problem-evaluator
- Escalate product ambiguity to pm-analyst
- Escalate implementation-level issues to backend
Arguments:
- feature-name: short identifier for the feature or module being triaged (required)
- scope-reference: module, service or component under investigation (required)
- problem-description: description of the problem or observed symptom (optional)
- constraints: optional constraints, boundaries or specific areas to investigate (optional)