ronnasayd/aipp/context-map
Claude Code and agent skill by ronnasayd/aipp
- community
Checking that the SKILL.md file is still in the repository.
Generate a map of all files relevant to a task before making changes. Works from a short or vague task description: expands it into concrete search seeds, locates code with rag-rat / codebase-memory / serena, traces dependencies and tests, and outputs a reviewed context map. Use when user says 'context map', 'map the files for this task', 'what files will this change touch', or before any non-trivial change.
Install
npx vskill@latest install ronnasayd/aipp/context-mapSKILL.md
The same SKILL.md was published earlier as ronnasayd/aipp/context-map.
Context Map
Before implementing any changes, map the files the task touches. Input may be one vague sentence: expand it into seeds, find anchors (code the task most directly changes), then expand outward from them. Every path in the map must be verified; absence of search results is not proof of absence.
Task
$ARGUMENTS
This skill runs as an isolated subagent: it has no conversation history and cannot ask the user anything. If the task above is empty, stop and return status: partial with reason: missing task description. If it names an earlier map, read it and extend it.
Workflow
Tool per question: tool-selection.md. Never cat whole files — outline first, read excerpts. Run independent searches in parallel.
rag-rat, codebase-memory and serena are servers behind mcp-manager: get_tool_schema(server, tool_name), then call_tool(server, tool_name, arguments). Use them first; grep alone misses callers and blast radius. If a server is unavailable, name it under Not Checked.
| # | Phase | Action | Gate | ||||
|---|---|---|---|---|---|---|---|
| 0 | Expand task | No tools. Write down: nouns → candidate symbols; verbs → behaviors; synonyms/naming variants (snake/camel, plural, abbrev); layer guess (UI/API/service/data/config/infra/docs); task type (feature/fix/refactor/config/docs); unknowns | seeds listed before any search | ||||
| 1 | Memory + orientation | Parallel: memory_query/memory_search on top 2-3 seeds (prior decisions, gotchas); repo_brief/get_architecture (scope with path if obvious); read AGENTS.md/CLAUDE.md rules for touched file types | constraints noted | ||||
| 2 | Locate anchors | semantic_search/search_graph per concept seed (parallel); symbol_lookup/find_symbol per exact name; grep literals (config keys, error messages, CLI flags, env vars) | 1-5 anchors; confirmed = hit by 2+ independent searches, else "uncertain" | ||||
| 3 | Handle ambiguity | Only when anchors split across unrelated areas, zero anchors after synonyms, or task type changes map materially: write an Open questions: list offering found candidates as options, and return status: partial with reason: open questions once the map is done. Otherwise state the assumption and continue | no stalling on answerable defaults | ||||
| 4 | Expand from anchors | Per anchor, parallel: dependents (find_callers/trace_path inbound, depth 2); dependencies (trace_callees/outbound, depth 1-2, keep only contracts the change could break); impact_surface on changed signatures; outline files for exact functions; string-level consumers; implementations | public/exported/cross-module hits flagged; depth cut-off noted | ||||
| 5 | Tests | find_callers/trace_path inbound with include_tests: true; name/path match (test_<mod>, <mod>.test.*, __tests__/, tests/); grep anchor names under test dirs | none found → record "no coverage" as risk, never invent a test | ||||
| 6 | Reference patterns | 1-3 existing implementations of the same kind of change: semantic_search on the pattern ("another hook that blocks on lint failure"), not the domain; search_graph by same label/decorator/base; query_graph for "all X that do Y"; prefer recent, most-imported | patterns cited by path | ||||
| 7 | Verify + assemble | check_index_coverage or direct outline on every cited path (drop or mark "unverified"); grep anchor names/config keys in `*.json\ | yaml\ | toml\ | md` (indexers skip them); fill risks from evidence; list Assumptions + Not Checked; fill output-template.md; run validator | validator exits 0; Risk Assessment evidence-based |
Phase 4 detail
- String-level consumers: the call graph misses tool names, config keys, CLI names referenced as strings (hooks, prompts, docs,
pyproject.tomlentry points).search_code/grepthe anchor name, modefiles. - Implementations/overrides:
find_implementationswhen the anchor is an interface, base class, or hook. - Stop at depth 2 unless blast radius is still growing.
Output
Fill output-template.md: Task Interpretation, Files to Modify, Dependencies, Test Files, Reference Patterns, Prior Decisions / Gotchas, Risk Assessment, Not Checked. ≤ 10 rows per table, ranked by impact.
Validation
Save the map (e.g. context-map.md) and run the gate before handing it over. Fix every ERROR; fix or justify every WARN.
Checks: required sections, exact table columns, no placeholder rows, valid Task type: + Anchors:, checkbox risks, cited paths exist. Exit 0 pass, 1 violations, 2 usage error.
Do not proceed with implementation until this map is reviewed.
Final response
Your final response must contain the complete map verbatim, the Open questions: list if any, and the saved path, alongside any other fields your final response requires.
Reference files
| File | Contents | |
|---|---|---|
| tool-selection.md | question → rag-rat / codebase-memory / serena tool table + rules | |
| output-template.md | verbatim map template the validator enforces | |
| validate_context_map.py | deterministic structure check for the finished map |
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/context-map/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/ronnasayd/aipp/context-map)