菜单
Use while debugging or troubleshooting failures, flaky tests, incidents, production alerts, CI failures, local dev setup issues, repeated bugs, prior fixes, failed attempts, and verification history.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when an agent needs recent or historical change context: what changed recently, regressions, merged PRs, tickets, docs, incidents, deployments, releases, and source-history ingestion.
Use when the task is centered on running, explaining, configuring, or troubleshooting the `potpie` command: doctor, login, pot management, source registration, search, graph workbench reads/writes, and pot scope behavior.
Use when the task can read or write the project-memory graph through the potpie CLI: discover the contract with `graph catalog`, read named views with `graph read`, resolve entity identity with `graph search-entities`, create validated plans with `graph propose`, commit plans with `graph commit --verify`, inspect quality with `graph quality`, or capture uncertain work with `graph inbox`. Also covers writing retrieval-grade descriptions and responding to nudges. Prefer this over the MCP context_* tools whenever the shell is available.
Use for project infra and architecture context: environments, adapters, runtime configuration, deployments, service dependencies, datastores, API contracts, ownership, incidents, and dependency blast radius.
Use before writing, modifying, reviewing, refactoring, or testing code so repo/project preferences surface: error handling, file structure, frameworks, logging, dependency choices, testing, security, API style, and naming. Also use after code work when a reusable project preference should be recorded.
Use when establishing, refreshing, or deeply understanding a repository's baseline memory in Potpie: purpose, application type, features, services/modules, environments, deploy shape, dependencies, API contracts, datastores, integrations, ownership, and explicit preferences. The harness reads authored and code-adjacent sources, then writes graph workbench mutations.
| name | potpie-debug-memory |
| description | Use while debugging or troubleshooting failures, flaky tests, incidents, production alerts, CI failures, local dev setup issues, repeated bugs, prior fixes, failed attempts, and verification history. |
Use this skill before digging into a failure so prior symptoms, known fixes, and failed attempts can guide the investigation.
Search by symptom first, not just component name. Include exact error text, commands, failing tests, environment, service, dependency, and synonyms.
potpie graph read \
--subgraph debugging \
--view prior_occurrences \
--query "<expanded symptom query>" \
--scope service:<service-name> \
--limit 12
If no service is known, omit --scope. If the failure smells like a regression,
correlate with the timeline:
potpie graph read \
--subgraph recent_changes \
--view timeline \
--format table \
--time-window 7d \
--query "<symptom feature dependency>" \
--limit 20
If dependencies, adapters, environments, or deploys matter, read infra too:
potpie graph read \
--subgraph infra_topology \
--view service_neighborhood \
--scope service:<service-name> \
--depth 2 \
--direction both
Treat prior fixes as leads. Check whether the same symptom, environment, version, dependency, data shape, command, or test path matches this incident. Failed prior attempts are useful because they prevent repeated work.
Record after the investigation when the learning is reusable: bug pattern, fix, verification, failed attempt, incident summary, runbook note, or setup gotcha.
Use the workbench write flow:
potpie --json graph catalog --task "record bug fix"
potpie graph search-entities "<service or symptom>" --limit 10
potpie --json graph describe debugging --view prior_occurrences --examples
potpie --json graph propose --file mutation.json
potpie --json graph commit <plan_id> --verify
potpie --json graph history --plan <plan_id>
Good debug memory includes the distinctive error text, repro signal, root cause or uncertainty, fix steps, verification status, scope, truth class, evidence, and a retrieval-grade description with symptom synonyms.
If the source may matter but the canonical update is uncertain, use
potpie --json graph inbox add instead of committing a weak fact.
Debug memory is harness-led: investigate and verify before writing. Do not use scanner-driven graph updates or record a bug/fix from filenames or logs alone.
Use this only when the potpie CLI is unavailable:
{"intent":"debugging","include":["prior_bugs","infra_topology","timeline"],"mode":"fast","source_policy":"references_only"}
Include families belong to the envelope surface (context_* MCP tools and
potpie resolve/potpie search); in the graph workbench they are served by the
graph views debugging.prior_occurrences, infra_topology.service_neighborhood,
and recent_changes.timeline — graph read does not accept include family
names.