DexForce/EmbodiChain / .agents/skills/project-dev-context
project-dev-context
VerifiedRoute EmbodiChain project-context and codebase-navigation requests through agent_context/MAP.yaml. Use to locate files, configs, defaults, entry points, registration paths or change sites; explain resolution chains; read, refresh or add project context. Chinese triggers include 文件在哪里、配置或默认值在哪里、 入口或注册逻辑在哪里、应该修改哪个文件、参考项目上下文、刷新项目上下文。
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.5 KB |
| agents/openai.yaml | config | 108 B |
| references/context-system.md | markdown | 3.3 KB |
| scripts/context.py | script | 26.7 KB |
SKILL.md
View on GitHub---
name: project-dev-context
description: >
Route EmbodiChain project-context and codebase-navigation requests through
agent_context/MAP.yaml. Use to locate files, configs, defaults, entry points,
registration paths or change sites; explain resolution chains; read, refresh
or add project context. Chinese triggers include 文件在哪里、配置或默认值在哪里、
入口或注册逻辑在哪里、应该修改哪个文件、参考项目上下文、刷新项目上下文。
---
# Project context and navigation
## Select and route
Read `agent_context/MAP.yaml` first. Select **navigate** (current code facts),
**read** (context), **refresh** (update a topic), or **add** (register a topic).
Specialized skills such as `/add-robot` own implementation; use this skill for
orientation and context maintenance.
Use the deterministic helper when matching is unclear or needs verification:
```bash
python .agents/skills/project-dev-context/scripts/context.py route 'QUERY'
```
Match a complete topic `id` first, including an explicitly named id within a
request; otherwise match aliases, then keywords. ASCII tokens have boundaries,
so `sim` does not match `simready`; Chinese phrases support substring matching.
Within the first matching alias/keyword tier, prefer the longest matching
phrase, then additional distinct matches. Equal best matches remain candidates;
never silently choose the first topic in MAP.
For ambiguity, use the request's subsystem/owning path to narrow candidates,
or inspect the specific symbol in their source files. If that still leaves
different answers, ask one concise clarifying question. For example,
`SceneManifest` belongs to both visualization and Task Program semantics;
`visualization SceneManifest` identifies the visualization topic.
## Read progressively
1. Load only the selected topic's `paths` (the overview).
2. Follow only the overview's detail links relevant to the question.
`related_topics` is navigation metadata, not an automatic load list.
3. For navigation, inspect the relevant symbols in `source_of_truth`, expanding
to callees/tests as needed. A listed directory is a search scope, not a
request to read every file. Current code takes precedence over context prose.
4. If nothing matches, use `rg --files` and `rg -n` for symbols, flags, config
keys, registries and imports. CLI ownership starts in `embodichain/cli/` and
`pyproject.toml`; `embodichain/__main__.py` is an entry wrapper.
5. Do not read `docs/source/` unless the user asks for Sphinx documentation.
Do not load writing conventions, this skill's reference schema or agent UI
metadata for ordinary reads/navigation. Answer with the owning entry point,
resolution path, recommended change site and focused validation when relevant.
## Maintain context
For refresh/add, read `references/context-system.md` and the conventions in
`defaults.write_contexts`. Use current implementation as evidence, not old notes.
Keep one owner for each contract and link from other topics. Preserve stable
topic ids and overview paths; linked detail pages carry optional depth.
Find topics potentially affected by a code change:
```bash
python .agents/skills/project-dev-context/scripts/context.py affected --base origin/main --explain
python .agents/skills/project-dev-context/scripts/context.py affected embodichain/lab/gym/envs/base_env.py
```
Review affected topics in the same change; this is an impact hint, not proof
that prose is stale or that unchanged paths are fresh. Choose **no update**,
**revise**, or **add** using the loaded maintenance conventions, and briefly state
the outcome in the PR or task result. Edit only guidance or metadata made
inaccurate or materially incomplete. Replace/consolidate the owning passage;
do not append a PR summary. Add a topic for a requested or recurring missing
domain, not automatically for every unmatched lookup.
After editing:
```bash
python .agents/skills/project-dev-context/scripts/context.py check
python .agents/skills/project-dev-context/scripts/context.py stats --base origin/main
python -m pytest -q -c /dev/null --noconftest tests/test_agent_context_map.py tests/test_agent_context_tools.py
```
Validate representative natural-language routes before and after routing changes,
including ambiguous and unmatched requests. For content restructuring, follow
representative questions from overview to owner and check how much unrelated
content must be read. Size/growth statistics are advisory, not a length gate.
Thin adapters only point here;
they need changes only when their local entry hints change.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |