workbench-l2-pressure-gate
Workbench L2 Pressure Gate
Use this skill when an issue asks for HarnessMax, remote evolution, leaderboard pressure, Research Vault grounding, long-running Hermes synthesis, or autonomous review sweeps that should improve the workbench over time.
L2 Pressure is the workbench's memory-pressure layer. Before choosing a path, the owner must pull relevant prior failures, constraints, and proven patterns from Research Vault or the closest available durable memory source. The result is a better next action, not a larger prompt.
Activation
Use this gate when any of these appear in the issue:
L2_PRESSURE: yesRV_PRESSURE: required- HarnessMax, remote evolve, remote VM, remote Hermes, leaderboard, or constant-evolve language
- a task that could repeat prior mistakes unless durable memory is consulted
Required Block
Post this before routing or implementation:
RV_PRESSURE_CHECK
objective:
owner:
vault_source:
queries_or_indexes_checked:
relevant_prior_failures:
proven_patterns:
l2_pressure_applied:
not_applied_and_why:
next_best_action:
verdict: PASS | FLAG | BLOCK
Source Order
Use the best available source, in this order:
- Remote Research Vault MCP endpoint if the issue proves it is available.
- Project-bound Research Vault repo/resource if attached to the issue.
- Local Research Vault MCP or filesystem source when the runtime is local.
- Workbench memory files,
SYNTHESIS.md,DECISIONS.md, and issue evidence when Research Vault access is unavailable.
If using a fallback, mark the result FLAG unless the task only needs
workbench-local memory.
Remote Rules
- Remote runtimes must not treat laptop
file://paths as valid Research Vault access. - Remote Research Vault MCP starts read-only:
vault_status,vault_search,vault_taxonomy, andvault_getonly. - Do not enable write, ingest, delete, or maintenance tools without a separate issue, explicit approval, and Supervisor review.
- Store only compact summaries in issue comments. Do not paste raw vault entries, secrets, OAuth material, request payloads, screenshots, or private logs.
- If MCP auth, endpoint, or routing fails, create or route to an RV MCP preflight issue instead of inventing memory.
Pressure Rules
- Convert memory into constraints and next actions.
- Prefer proven failure modes over generic advice.
- Reject novelty when a prior pattern already identifies the highest-yield path.
- Name the route not taken when leaderboard pressure would tempt broad work.
- Escalate to Supervisor when memory conflicts with the issue objective.
Closeout
For L2 Pressure tasks, PASS requires:
- the required block was posted;
- at least one durable source was checked or a fallback was justified;
- prior failures or proven patterns changed the plan, review, or route; and
- the final report states whether the pressure improved, blocked, or merely confirmed the current path.
If no relevant memory exists, report PASS only when the search was specific and
bounded. Broad "nothing found" claims are FLAG.
More from fearvox/multica-ultimate-workbench
workbench-conductor
Two-ring orchestration, routing, issue and comment discipline, and role boundaries for the Multica Workbench.
5workbench-sdd
Specification-driven development from raw requirement to product design, technical design, task list, execution, and verification.
5workbench-self-awareness-infra
Capability discovery and current-state verification for Heavy Path, ambiguous repo/runtime ownership, and runtime-dependent Standard Path work.
5workbench-design-docs
Product design, technical design documents, user-facing copy, specs, diagrams, and handoff documentation.
5workbench-token-context-discipline
Compact context, cache-aware execution, scoped evidence reads, and role-specific skill attachment discipline.
4workbench-product-brainstorming
Bounded product ideation, workflow design, ambition checks, tradeoffs, and smallest-test shaping.
4