بنقرة واحدة
potpie-context
// Use when an agent needs project context from the context engine through MCP tools. Covers context_resolve recipes for feature, debugging, review, operations, docs, and onboarding without adding separate context tools.
// Use when an agent needs project context from the context engine through MCP tools. Covers context_resolve recipes for feature, debugging, review, operations, docs, and onboarding without adding separate context tools.
Use when the task involves running, explaining, or troubleshooting the Potpie CLI. Covers doctor, login, pot management, search, and ingest commands.
Use when the Potpie CLI is failing — especially around API key, URL config, 401, 404, pot scope, or search/ingest HTTP errors.
Use when the task is about pot resolution: pot use, env maps, git origin parsing, repo-to-pot mapping, or search/ingest scope inference.
| name | potpie-context |
| description | Use when an agent needs project context from the context engine through MCP tools. Covers context_resolve recipes for feature, debugging, review, operations, docs, and onboarding without adding separate context tools. |
Use this skill when the task requires gathering, verifying, or recording project context through the Potpie MCP tools.
Use only the minimal context port:
context_resolve — primary task context wrap.context_status — cheap pot readiness, freshness gaps, and recommended recipe.context_search — narrow follow-up lookup after context_resolve.context_record — durable learnings: decisions, fixes, preferences, workflows, feature notes, doc references, incident summaries.Do not request or invent separate tools for feature context, debugging context, operations context, source lookup, or docs context. Express those through context_resolve parameters.
pot_id.scope from known repo, branch, files, services, features, environment, PR, ticket, user, or source refs.context_status when setup or readiness is uncertain or the task is broad.context_resolve with the matching recipe.coverage, freshness, quality, fallbacks, open_conflicts, and source_refs before relying on the answer.source_policy="summary" for compact source-backed summaries.source_policy="verify" and mode="verify" before production-impacting or disputed facts.source_policy="snippets" or mode="deep" only for bounded source detail within budget.context_record when the work discovers reusable project memory.Feature work:
{"intent":"feature","include":["purpose","feature_map","service_map","docs","tickets","decisions","recent_changes","owners","preferences","source_status"],"mode":"fast","source_policy":"references_only"}
Debugging:
{"intent":"debugging","include":["prior_fixes","diagnostic_signals","incidents","alerts","recent_changes","config","deployments","owners","source_status"],"mode":"fast","source_policy":"references_only"}
Review:
{"intent":"review","include":["artifact","discussions","owners","recent_changes","decisions","preferences","source_status"],"mode":"balanced","source_policy":"summary"}
Operations:
{"intent":"operations","include":["deployments","runbooks","alerts","incidents","scripts","config","owners","source_status"],"mode":"balanced","source_policy":"summary"}
Docs:
{"intent":"docs","include":["docs","decisions","source_status"],"mode":"fast","source_policy":"references_only"}
Onboarding:
{"intent":"onboarding","include":["purpose","repo_map","service_map","docs","local_workflows","agent_instructions","source_status"],"mode":"fast","source_policy":"references_only"}
Use context_record with record_type one of:
decision — architectural or product decision.fix — concrete fix that resolved a bug or incident.bug_pattern — recurring symptom/root-cause pattern.investigation — useful diagnostic path or ruled-out hypotheses.diagnostic_signal — log line, metric, alert, or error signature.preference — team/user/project preference.workflow — local, release, deployment, or operational workflow.feature_note, service_note, runbook_note, integration_note, doc_reference, incident_summary.Keep records compact and source-reference-first.
quality.status=good — graph context acceptable for low-risk orientation.quality.status=watch — verify stale or unverified facts before high-impact action.quality.status=degraded — prefer source truth; consider recording an alignment correction.quality.recommended_maintenance when the response suggests jobs such as verify_entity, refresh_scope, resync_source_scope, repair_code_bridges, or expire_stale_facts.