steelmorgan/v8-session-manager
framework_eng/skills/tool-usage/v8-session-manager/SKILL.md
Use for working with the 1С session manager: launching, configuration, connecting clients, reading session_list, calling proxied MCP tools from 1С extensions. Helps with errors «no active sessions» / «session_id required» and connecting a client via `mcpMode=ws`.
$ install --global
skillsauthnpx skillsauth add steelmorgan/1c-agent-based-dev-framework v8-session-managerInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Jun 13, 2026, 6:01 AM161.4s6 files scanned
SKILL.md
- name:
- v8-session-manager
- description:
- Use for working with the 1С session manager: launching, configuration, connecting clients, reading session_list, calling proxied MCP tools from 1С extensions. Helps with errors «no active sessions» / «session_id required» and connecting a client via `mcpMode=ws`.
- # WARNING:
- their names in tools/list are read from the persistent tools-cache
- # tool error `isError:
- true, _meta.error_code="no_live_session"`.
v8-session-manager
A thin MCP aggregator: accepts WS connections from 1С clients and publishes their MCP tools on a single HTTP endpoint for the AI agent.
What the manager provides itself
| Capability | Source |
|---|---|
| Built-in tool — session_list (read-only registry snapshot) | manager |
| Built-in tool — tools_cache_reset (full reset or by config_id) | manager (ADR-0035) |
| Showcase of proxied tools from connected clients | 1С extensions |
| Persistent showcase cache (workPath/tools_cache.json, TTL 5d) | manager (ADR-0035) |
| Routing a call to the correct session by session_id | manager |
| Soft reconnect of a client by client_uid | manager |
| FIFO order of calls into one session | manager |
Everything else (domain tools — form descriptions, test runs, navigation, etc.) is added by 1С extensions, not by the manager. There is a separate skill for each extension.
Proxied tools cache (ADR-0035) — key point
The manager's tools/list is read from a persistent cache on disk, not only from live WS sessions. Implications for the agent:
- A tool name in
tools/list≠ a successful call. The cache survives client disconnect and manager restart — the name stays on the showcase, but a call without a live session returns an MCP tool errorisError:true, _meta.error_code="no_live_session". This is not a bug, it is the contract. - Why it is done this way: some MCP harnesses (in particular Claude Code) respond unreliably to
notifications/tools/list_changed. The persistent cache removes the dependency on stable notification handling. - When
tools_cache_resetis needed: when a tool has been intentionally removed from the extension and will not come back (or the configuration has been removed completely). Otherwise it will linger until the TTL expires (by default 5 days from the lastsession.register). Full reset — without arguments; targeted —{"config_id": "<id>"}(taken fromsession_list[*].config_id). - What the cache does NOT do: it does not start 1С, does not reproduce the tool response, does not replace a live session. It only stores names and
inputSchema.
Details — references/sessions-and-tools.md § «Persistent cache and tools_cache_reset».
Boundaries
The manager does not:
- launch 1С clients (that is the job of an external orchestrator, typically
v8-runner); - store state across restarts (the registry is in-memory);
- contain business logic (transport + routing only);
- manage the infobase.
Task Routing
| Task | Reference |
|---|---|
| What each layer of the stack does (addin → devkit → BSL → manager → AI) | references/architecture.md |
| Bring up the manager, connect a 1С client | references/bootstrap.md |
| Read session_list, call a tool, understand why it is missing | references/sessions-and-tools.md |
| Add a new tool to a 1С extension | references/extending-tools.md |
| Manager does not start / client is not visible / tool is hidden / call fails | references/troubleshooting.md |
Guardrails (hard)
- Do not edit the manager source code (
src/,Cargo.toml,systemd/,etc/,spec/, ADR indocs/decisions/) — this is the upstream repository. All manager-level changes are coordinated in a separate task. - Do not create or modify MCP tools without the user's direct permission. Tools live in 1С extensions (
exts/<extension>/); editing/adding them means changing the public contract. - Do not pull business logic into the manager. If a task requires "the manager should do X", that is a signal that X belongs either in the extension or in the launch orchestrator.
- Do not try to start a 1С client through the manager. The manager only accepts an incoming WS connection. Starting 1С is the responsibility of
v8-runner. - Ask the user before building/restarting the client. Any operation that changes project state (build, restart) requires confirmation.
Related Skills
steelmorgan/source-of-truth
testing
VerifiedTrustedCommunity
MUST use BEFORE making a judgment about the cause of a conflict, a test failure, or an artifact dispute. Defines the end-to-end verification method L1→L6 and the classification of the first broken link.
69SKILL.mdUpdated Jun 13, 2026
steelmorgan/skill-learning
development
VerifiedTrustedCommunity
MUST use AFTER a work cycle with ≥2 iterations (wrote → error → fixed → success). Provides the retrospective procedure and the format for recording practice/anti-patterns in references/learned-patterns.md or {project}/.context/learned-patterns.md.
69SKILL.mdUpdated Jun 13, 2026
steelmorgan/rlm-workflow
tools
VerifiedTrustedCommunity
MUST use WHEN you are writing reusable knowledge into RLM (pattern / architectural decision / stable domain fact) OR reading it before a non-trivial task/solution in the domain. Provides the breakdown of native-push vs RLM-pull, tools for writing and reading RLM, H-MEM levels, and hygiene.
69SKILL.mdUpdated Jun 13, 2026
steelmorgan/quick-fix
testing
VerifiedTrustedCommunity
MUST use WHEN the task is classified as simple (< 20 lines, 1 file, no new metadata objects, no architectural decisions). Provides a short cycle of 3 steps with a guard on the self path and mandatory verify.
69SKILL.mdUpdated Jun 13, 2026