feat(self-audit): a self_audit recipe carries the brief continuous_improvement never had
deploy / test (push) Successful in 4m55s
deploy / build (push) Successful in 1m54s

Role system_prompts are inert on mission turns, so a team without a recipe has
no default brief and every caller must write one (v2 wrote the wrong one: 0/50
calls read the record). The recipe points the phase at
/mission/memory/PROJECT-MEMORY.md, requires verbatim quotes only where a
pattern is reported, and names the clean-record and no-record outcomes as
passes — the v3 condition failed a correct clean audit for paraphrasing.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
This commit is contained in:
Omar Sobh
2026-09-22 17:25:36 -05:00
co-authored by Claude Opus 5.5
parent e789f6f7ea
commit fca828b5a1
5 changed files with 50 additions and 0 deletions
@@ -91,6 +91,7 @@ const TEMPLATE_LABEL: Record<TemplateKind, string> = {
refactor: "Refactor",
benchmark: "Benchmark",
continuous_research: "Continuous Research",
self_audit: "Self-audit",
custom: "Custom",
};
@@ -36,6 +36,7 @@ const TEMPLATE_BADGE: Record<TemplateKind, { label: string; color: string }> = {
refactor: { label: "REFACTOR", color: "#ffb44a" },
benchmark: { label: "BENCHMARK", color: "#83e6a5" },
continuous_research: { label: "PODCAST", color: "#e0b0ff" },
self_audit: { label: "AUDIT", color: "#b8c4a0" },
custom: { label: "CUSTOM", color: "#8a8a92" },
};
+1
View File
@@ -91,6 +91,7 @@ const PALETTES: Record<string, Palette> = {
security_hardening: SECURITY,
refactor: REFACTOR,
benchmark: BENCHMARK,
self_audit: RESEARCH,
custom: BASE,
};
+1
View File
@@ -9,6 +9,7 @@ export type TemplateKind =
| "refactor"
| "benchmark"
| "continuous_research"
| "self_audit"
| "custom";
export type MissionStatus =