Enforces that the agent queries observability tools (VictoriaLogs, VictoriaMetrics, VictoriaTraces) during development instead of guessing at runtime behavior. Triggers when the agent is debugging, investigating errors, or verifying runtime behavior.
From codeharnessnpx claudepluginhub ivintik/private-claude-marketplace --plugin codeharnessThis skill uses the workspace's default tool permissions.
references/observability-querying.mdreferences/otlp-instrumentation.mdSearches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Guides implementation of event-driven hooks in Claude Code plugins using prompt-based validation and bash commands for PreToolUse, Stop, and session events.
The agent must SEE what the code does at runtime. No guessing. Query logs, metrics, and traces.
Query observability endpoints in these situations:
Before querying, follow this decision flow:
Check if Docker stack is healthy:
codeharness status --check-docker
If unhealthy, the agent cannot query. Start the stack first:
codeharness stack start
Query the appropriate endpoint:
localhost:9428localhost:8428localhost:16686See knowledge/observability-querying.md for full query patterns and endpoint formats.
After querying, act on results:
After querying logs, record that the query happened:
codeharness state set session_flags.logs_queried true
This session flag is checked by quality gates before commits. It resets to false at the start of each session (via session-start.sh hook).
Observability is always enabled (mandatory). If endpoints are unreachable:
codeharness stack startcodeharness init --otel-endpoint <url>codeharness status --check-dockerknowledge/observability-querying.mdknowledge/otlp-instrumentation.md