Menu
Maintainer-only โ diagnose CodeClone's OWN runtime via Platform Observability after explicit observer enable. Not a user-repo quality signal.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Human repository setup readiness โ status, plan, apply, and wizard over the CLI (not MCP).
Rank demonstrated architectural problems from one canonical run plus targeted source verification. Read-only; priorities are response-local heuristics, never CodeClone findings.
Inspect structural blast radius before editing โ dependents, clone cohort, risk signals, do-not-touch boundaries. Read-only; does not declare intent.
MANDATORY HARD GATE when CodeClone MCP is connected. Before ANY repository file write, obtain edit_allowed=true from start_controlled_change. Before claiming done/verified/ready, finish_controlled_change must accept and clear the intent.
Use CodeClone Engineering Memory through MCP to retrieve scoped evidence before edits and preserve durable incidents, decisions, risks, contradictions, and verification anchors as draft candidates. Memory never grants edit permission; agents propose, humans govern.
Before broad rg/grep, non-trivial planning, unclear scope, contract/schema work, or resuming another agent โ call get_implementation_context once to bound the implementation frontier. Then use targeted source reads. Read-only; never grants edit permission.
| name | codeclone-platform-observability |
| description | Maintainer-only โ diagnose CodeClone's OWN runtime via Platform Observability after explicit observer enable. Not a user-repo quality signal. |
Use ONLY when developing CodeClone itself (instrumentation, MCP server, memory pipelines, projection workers, observer storage/rendering). It explains CodeClone's execution โ NOT the analyzed project's code quality. High DB activity = CodeClone ran SQL, not a user-repo defect.
| Audience | Use? |
|---|---|
| CodeClone maintainers / contributors | yes โ after explicit observer setup |
| Users analyzing THEIR repo | no โ use codeclone-review / codeclone-hotspots / change control |
| Repo quality / CI / gate questions | no โ observer data is not analysis truth |
Disabled by default โ the tool returns status=disabled / no_store (no error, no
behavior change). Before any diagnostic:
export CODECLONE_OBSERVABILITY_ENABLED=1
Restart the SAME process that produces telemetry (CLI, codeclone-mcp, or projection
worker). A store appears only after one instrumented op completes:
.codeclone/db/platform_observability.sqlite3. There is no
[tool.codeclone.observability] pyproject table โ env only.
help(topic="observability") # contract + anti-patterns (optional)
โ reproduce with CODECLONE_OBSERVABILITY_ENABLED=1
โ query_platform_observability(section="summary", window="latest")
โ follow recommended_next_sections, ONE section per call
Sections:
summary | slow_operations | mcp_tool_matrix | db_cost | memory_pipeline_cost | correlated_chains | costly_noops | pipeline | agent_context.
Params: absolute root, detail_level=compact|normal, limit 1โ50, window=latest
or a correlation id.
Human cockpit (not MCP): codeclone observability trace --root . --last 50 --html /tmp/x.html.
Key / easily-misread fields; the real response carries more.
| Field | Meaning |
|---|---|
status | ok / disabled / no_store (stop โ verify env + reproducer process; do not retry blindly) |
recommended_next_sections | follow these, one per call |
row verdict | query_chatty / context_heavy / ok โ perf signal, NOT a code-quality verdict |
costly_noops | redundant-work hints |
affects_analysis_truth / affects_edit_permission | always false โ observer never gates or authorizes |
MCP payload context metrics are estimated context units. For responses carrying
context_governance, the observer uses that envelope's estimated value; do not
describe it as an exact model tokenizer count.
tests/test_observability_*.py.