📖About

Three gaps found auditing the new service-track spending limits.

1. Carol falls silent instead of apologising. Her message handler recognises only the wording the RETIRED lane limit used. The live limit refuses with different wording, so when her track is spent she raises an error instead of sending her 'reached my budget for today' reply — a real person on WhatsApp gets no answer. Every other consumer of a refusal already matches the whole family; only the one talking to a human does not.

2. Thinking and chatting do not declare themselves. A ticket's kind is derived from the droid's process type, so every scheduled droid says 'sweep' — including the wake cycles. Elrond's thinking is currently labelled a sweep. It reaches the right track only because a safety net recognises the droid by name. The operator's ruling is that the ACTIVITY decides the track, so the declared kind must say what the work is.

3. The modules that now enforce every limit are unprotected while the retired ones are locked. The live spending control, the media spending gate, the work ticket and the paid-call fence are all writable by any autonomous lane, whereas the four superseded budget modules are root-owned and kernel-locked. The protection is on the wrong files.

⚖️Decisions

• Elrond's bypass methodology checklist (a reminder, not a gate -- you've got this): 0. File it requested_mode='bypass' (planner-vs-bypass is a deliberate choice). bypass_start REFUSES a non-bypass initiative (CAROL-INI-1846), and the dispatcher only skips the bypass lane when the mode says bypass -- a 'planner' mistag lets Merlin's pipeline grab the placeholder step and block your finished work. 1. Filed as planned status -- let the bypass claim/activate it; never file active. 2. Open the bypass (bypass_start) with your droid id + the remediation answer (remediates_initiative_id=NNN, or remediates_nothing=True). 3. Work the blocks for your work-type: template -> design -> code -> test -> review. Do the real work; record decisions on the initiative as you make them. 4. Reality is recorded for you at close -- code (files changed), each decision, and the twin-review verdict become real activities tied to this initiative and show in the Activity Tracker like a planner run (CAROL-INI-1840). No dummy rows. 5. Keep the initiative status moving; it parks in 'reviewing' and is tagged uat-pending for you at close (CAROL-INI-1836), so the stuck-watchdog leaves it alone until UAT. 6. Close runs the gates (design/architecture compliance + caller-audit). If a gate flags something pre-existing or unrelated to your change, waive it with a clear written rationale -- audit, don't skip. 7. Bypass skips the planner's auto-orchestration, NOT the standards. Same template checklist, same review, same observability as a planner run. (elrond)
• [status-router] planned -> executing | event=bypass_executing | bypass transition (or-bx-01)
• Carol degrades gracefully on ANY refuse-before-spend marker, not one prefix — Her handler matched only LANE_BUDGET_PARKED, which the retired lane gate produced and the live gate never emits. Matching the family (BUDGET_PARKED) is what every other consumer already does. A prefix match is how a refusal stopped reaching the only agent that talks to a person. (orion)
• The declared activity is derived from the spend gate's own rule, not a second copy — The gate already decides which droids are thinking or chatting, because it must - it is the safety net for work that declares nothing. Inverting its answer keeps one rule; writing a parallel list in the ticket module would drift. (orion)
• Protection follows what governs money TODAY, and includes the ledger — The four locked budget modules are three-quarters retired as gates, while the live control, the media gate, the ticket and the paid-call fence were writable by any autonomous lane. The cost log is included because a limit is only as trustworthy as the ledger it is measured from - an actor that can rewrite the ledger can spend without appearing to. (orion)
• Those rows are marked unenforced and nothing reads them for enforcement, so it is harmless today - but it is a live process maintaining a dead control, and it is the same mechanism that would flatten the operator's deliberate 1/2/3/10 ladder if anyone re-pointed it. Filed rather than rushed: it has the widest blast radius and the least urgency, and two parallel sessions rewrote this area hours ago. (orion)
• Raising it is the operator's call and the ladder is his (1/2/3/4/5/10). (orion)
• [status-router] executing -> reviewing | event=bypass_reviewing | bypass transition (or-bx-01)
• [status-router] reviewing -> closed | event=operator_signoff | Auto-accepted (CAROL-INI-1859): Orion-initiated, >2 days in reviewing with no objection. (el-srac-01)

✅Success criteria

• When Carol's daily spending limit is reached she still replies to the person, telling them she has paused for the day, instead of failing silently. (must_have)
• A wake cycle's spending is declared as thinking, and a chat's as chatting, so the record says what the work was rather than calling it a sweep. (must_have)
• The modules that decide and enforce spending limits are listed as protected machinery, so an autonomous agent is refused when it tries to change them. (must_have)