How do I install the Rcc Dev Skills skill?

Run npx skillmds@latest add jasonzhangf/rcc-dev-skills in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Rcc Dev Skills safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Rcc Dev Skills?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Rcc Dev Skills free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Rcc Dev Skills?

jasonzhangf (@jasonzhangf) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Rcc Dev Skills by jasonzhangf — Agent Skill

Name: Rcc Dev Skills
Author: jasonzhangf

Rcc Dev Skills

RouteCodex V3 evidence-first development: query resource/function/mainline/verification maps, prove first divergence, patch one owner, then run mapped gates and installed same-entry replay. Use for debug, protocol, provider, routing, continuation, SSE, servertool, or runtime changes.

jasonzhangf 8ca8880 17 files · 23.8 KB Updated Sep 17, 2026

File contents

RouteCodex V3 Development

This skill owns V3 command and gate sequencing. Inherit general methods from coding-principals and boundaries from project AGENTS.md; do not also run a generic development phase workflow. Pure instruction/document edits validate affected references and contracts; runtime build/install/restart/replay applies only when runtime behavior or its delivery changes.

Command-First Flow

Read AGENTS.md.
Query maps before source:

rcc_task_feature='<feature_id>'
rg -n "feature_id: ${rcc_task_feature}" \
  docs/architecture/v3-resource-operation-map.yml \
  docs/architecture/v3-function-map.yml \
  docs/architecture/v3-mainline-call-map.yml \
  docs/architecture/v3-verification-map.yml

Lock resource edge, owner, allowed/forbidden paths, caller/callee, and gates. Missing or ambiguous binding blocks edits. Binding must agree with AGENTS.md; conflicts require remediation.
Read mapped source, generated review surface, current run notes, project MEMORY.md, and relevant history.
For defects, capture one request id and find first semantic divergence. Keep one active hypothesis.
Record red evidence: focused failing test, saved failing shape, or controlled replay.
Patch only mapped owner. Run positive and negative tests.
Run feature required_gates, then project architecture gate:

npm run verify:v3-architecture-ci

Runtime-impacting change: load references/50-rcc-config-ssot.md; prove build, install, config check, managed restart, all-listener health, and same-entry replay.
If a failing runtime sample exists, include its exact replay or same-entry semantic equivalent in step 9 evidence. Review only after verification; do not repeat an unchanged replay solely for this step.

Review Gate

After verification and before commit/merge, use the reviewer selected by global AGENTS.md and its shared review standards. Apply the project Semantic Invariants and mapped owner/edge/gate bindings. Do not duplicate the global checklist or historical-finding policy here.

Routes

Need	Reference
owner and gates	`references/40-owner-registry.md`
first-divergence pipe debug	`references/10-pipedebug-flow.md`
protocol/SSE/continuation	`references/25-protocol-sse-continuation-boundary.md`
config/install/runtime replay	`references/50-rcc-config-ssot.md`
servertool	V3 maps + `docs/agent-routing/30-servertool-lifecycle-routing.md`
error chain	`references/96-unified-error-path-audit.md`
selected provider model	`references/96-v3-selected-provider-model-binding-sop.md`
continuation cache	`references/97-continuation-cache-compliance.md`
provider request/error	`references/98-provider-request-dryrun-and-request-error-debug.md`

Report

Report goal, owner, first divergence, changed paths, red/green evidence, mapped gates, installed runtime evidence, review result, remaining gap, and next transition. State why edited owner is unique and which adjacent layers were ruled out.

---name: rcc-dev-skillsdescription: RouteCodex V3 evidence-first development: query resource/function/mainline/verification maps, prove first divergence, patch one owner, then run mapped gates and installed same-entry replay. Use for debug, protocol, provider, routing, continuation, SSE, servertool, or runtime changes.---# RouteCodex V3 DevelopmentThis skill owns V3 command and gate sequencing. Inherit general methods from `coding-principals` and boundaries from project `AGENTS.md`; do not also run a generic development phase workflow. Pure instruction/document edits validate affected references and contracts; runtime build/install/restart/replay applies only when runtime behavior or its delivery changes.## Command-First Flow1. Read `AGENTS.md`.2. Query maps before source:```bashrcc_task_feature='<feature_id>'rg -n "feature_id: ${rcc_task_feature}" \ docs/architecture/v3-resource-operation-map.yml \ docs/architecture/v3-function-map.yml \ docs/architecture/v3-mainline-call-map.yml \ docs/architecture/v3-verification-map.yml```3. Lock resource edge, owner, allowed/forbidden paths, caller/callee, and gates. Missing or ambiguous binding blocks edits. Binding must agree with `AGENTS.md`; conflicts require remediation.4. Read mapped source, generated review surface, current run notes, project `MEMORY.md`, and relevant history.5. For defects, capture one request id and find first semantic divergence. Keep one active hypothesis.6. Record red evidence: focused failing test, saved failing shape, or controlled replay.7. Patch only mapped owner. Run positive and negative tests.8. Run feature `required_gates`, then project architecture gate:```bashnpm run verify:v3-architecture-ci```9. Runtime-impacting change: load `references/50-rcc-config-ssot.md`; prove build, install, config check, managed restart, all-listener health, and same-entry replay.10. If a failing runtime sample exists, include its exact replay or same-entry semantic equivalent in step 9 evidence. Review only after verification; do not repeat an unchanged replay solely for this step.## Review GateAfter verification and before commit/merge, use the reviewer selected by globalAGENTS.md and its shared review standards. Apply the project Semantic Invariantsand mapped owner/edge/gate bindings. Do not duplicate the global checklist orhistorical-finding policy here.## Routes| Need | Reference || --- | --- || owner and gates | `references/40-owner-registry.md` || first-divergence pipe debug | `references/10-pipedebug-flow.md` || protocol/SSE/continuation | `references/25-protocol-sse-continuation-boundary.md` || config/install/runtime replay | `references/50-rcc-config-ssot.md` || servertool | V3 maps + `docs/agent-routing/30-servertool-lifecycle-routing.md` || error chain | `references/96-unified-error-path-audit.md` || selected provider model | `references/96-v3-selected-provider-model-binding-sop.md` || continuation cache | `references/97-continuation-cache-compliance.md` || provider request/error | `references/98-provider-request-dryrun-and-request-error-debug.md` |## ReportReport goal, owner, first divergence, changed paths, red/green evidence, mapped gates, installed runtime evidence, review result, remaining gap, and next transition. State why edited owner is unique and which adjacent layers were ruled out.

Frequently asked questions