opencode-swarm 7.114.6 → 7.114.8

package/.opencode/skills/engineering-conventions/SKILL.md

@@ -52,6 +52,39 @@ The OpenCode `test_runner` tool is for **targeted agent validation** with explic
 - For repo validation, run the shell commands in `contributing.md` / `TESTING.md` directly (per-file isolation loops + tier orchestration).
 - `scope: 'all'` is gated behind the `SWARM_ALLOW_FULL_SUITE=1` env var (intended for opt-in CI mirrors only); there is no `allow_full_suite` arg. Default to `files: [...]` instead.
 
+## Agent prompt strings — escaping pitfalls
+
+Agent prompts in `src/agents/*.ts` are large TypeScript template literals. They frequently contain characters that have special meaning inside template literals and cause silent parse errors if unescaped:
+
+| Character | Inside template literal | Correct escape |
+|-----------|------------------------|----------------|
+| Backtick `` ` `` | Terminates the literal | `` \` `` (single backslash — renders as `` ` `` in output) |
+| `${` | Starts an interpolation | `\${` (single backslash) |
+| Literal backslash `\` | Consumed by escape processing | `\\` (double backslash renders as `\` in output) |
+
+**The most common failure pattern:** A coder adds an inline code example containing backticks to an agent prompt string. The unescaped backtick silently terminates the template literal, producing a `SyntaxError: Unexpected identifier` or `Unexpected token` at the character *after* the backtick — which appears unrelated to the actual cause.
+
+```typescript
+// WRONG — unescaped backtick terminates the template literal
+const PROMPT = `
+Use `bun:test` for all tests.   // ← bare backtick before "bun" closes the literal
+`;
+
+// CORRECT — single backslash before each backtick; renders as Use `bun:test` in output
+const PROMPT = `
+Use \`bun:test\` for all tests.
+`;
+
+// OVER-ESCAPED (also wrong) — triple backslash produces literal \` in the rendered prompt
+const PROMPT = `
+Use \\\`bun:test\\\` for all tests.  // renders as: Use \`bun:test\` (backslashes visible)
+`;
+```
+
+**Detection:** If `bun run build` or `bun --smol test` reports a parse error at a line number that seems far from any recent change, search the surrounding lines for an unescaped backtick inside a template literal.
+
+**Prevention:** After adding any inline code example to an agent prompt, run `bun run build` immediately — the TypeScript compiler catches unescaped backticks as a syntax error before any tests run.
+
 ## The invariant-audit gate (PR-time)
 
 Every PR that touches a relevant area must include an `## Invariant audit` section in its description. The format is in `AGENTS.md` ("Invariant audit required in PRs"). The `commit-pr` skill enforces this gate before push/PR — load it before committing.

package/README.md

@@ -39,7 +39,7 @@ Most AI coding tools let one model write code and ask that same model whether th
 - 🔄 **Phase completion gates** — completion-verify and drift verifier gates enforced before phase completion
 - 🔁 **Resumable sessions** — all state saved to `.swarm/`; pick up any project any day
 - 🖥️ **PR Monitor** — GitHub PR subscription and background polling via `gh` CLI; delivers real-time CI, review, and merge status updates via the AutomationEventBus (FR-001, opt-in via `pr_monitor.enabled: true`). Subscribe with `/swarm pr subscribe <pr-url|owner/repo#N|N>`; unsubscribe with `/swarm pr unsubscribe <pr-url|owner/repo#N|N>`; check status with `/swarm pr status`. Enable `auto_pr_feedback: true` in `pr_monitor` config to inject `[MODE: PR_FEEDBACK pr="URL"]` on CI failures and merge conflicts automatically.
-- 🌐 **12 full language profiles** (TypeScript, Python, Go, Rust, Java, Kotlin, C/C++, C#, Ruby, Swift, Dart, PHP) with **tree-sitter parse validation across 20 grammars** (adds JavaScript, CSS, Bash, PowerShell, INI, Regex — and `.tsx` / `.c` aliases) — extending: see [docs/adding-a-language.md](docs/adding-a-language.md)
+- 🌐 **13 full language profiles** (TypeScript, JavaScript, Python, Go, Rust, Java, Kotlin, C/C++, C#, Ruby, Swift, Dart, PHP) with **tree-sitter parse validation across 20 grammars** (adds CSS, Bash, PowerShell, INI, Regex — and `.tsx` / `.c` aliases) — extending: see [docs/adding-a-language.md](docs/adding-a-language.md)
 - 🛡️ **Built-in security** — SAST, secrets scanning, dependency audit per task
 - 🔒 **Scope enforcement** — Validates write targets against declared scope with cross-process persistence, TTL expiry, and scope-aware destructive command blocking. **Handles both single-string and array-based path arguments** (`files[]`, `paths[]`, `targetFiles[]`) to prevent scope bypass via multi-file tool calls.
 - 📝 **Shell write detection** — Static analysis of POSIX/PowerShell/cmd commands to detect file writes (redirects, builtins, in-place editors, network downloads, archive extraction, git destructive ops) before execution
@@ -550,14 +550,14 @@ These quality gates run locally. No Docker is required. Config-gated features su
 
 ### Context Budget Guard
 
-The Context Budget Guard monitors how much context Swarm is injecting into the conversation. It helps prevent context overflow before it becomes a problem.
+The Context Budget Guard monitors how full the model's context window is getting across the whole conversation. It helps prevent context overflow before it becomes a problem.
 
 ### Default Behavior
 
 - **Enabled automatically** — No setup required. Swarm starts tracking context usage right away.
-- **What it measures** — Only the context that Swarm injects (plan, context, evidence, retrospectives). It does **not** count your chat history or the model's responses.
-- **Warning threshold (0.7 ratio)** — When swarm-injected context reaches ~2800 tokens (70% of 4000), the architect receives a one-time advisory warning. This is informational — execution continues normally.
-- **Critical threshold (0.9 ratio)** — When context reaches ~3600 tokens (90% of 4000), the architect receives a critical alert with a recommendation to run `/swarm handoff`. This is also one-time only.
+- **What it measures** — The estimated total tokens across **all** messages in the conversation (every text part of every message — your prompts, the model's responses, and injected content alike), divided by the model's context window. The window comes from `context_budget.model_limits` (default `128000`), resolved per model/provider. It does **not** measure only swarm-injected content, and it does **not** read `max_injection_tokens`.
+- **Warning threshold (0.7 ratio)** — When total conversation tokens reach 70% of the model context window (e.g. ~89,600 of 128,000), the architect receives a one-time advisory warning. This is informational — execution continues normally.
+- **Critical threshold (0.9 ratio)** — When total conversation tokens reach 90% of the model context window (e.g. ~115,200 of 128,000), the architect receives a critical alert with a recommendation to run `/swarm handoff`. This is also one-time only.
 - **Non-nagging** — Alerts fire once per session, not repeatedly. You won't be pestered every turn.
 - **Who sees warnings** — Only the architect receives these warnings. Other agents are unaware of the budget.
 
@@ -588,17 +588,18 @@ Swarm includes an intelligent skill propagation system that tracks, validates, a
 
 ```json
 {
-  "skill_propagation": {
+  "skillPropagation": {
     "enabled": true,
     "enforce": false,
-    "scoring": {
-      "threshold": 0.5,
-      "max_recommendations": 5
-    }
+    "audiences": []
   }
 }
 ```
 
+- **`enabled`** (boolean, default `true`) — turns skill-propagation scoring, recommendations, and warnings on or off. The mandatory explicit-skill-reference integrity check still runs even when this is `false`.
+- **`enforce`** (boolean, default `false`) — when `true`, blocks delegations that omit the `SKILLS:` field instead of only warning.
+- **`audiences`** (string array, default `[]`, max 16 entries, de-duplicated) — extra audience tokens to broaden which skills are considered in scope for this project, in addition to the active runner's own audience. Each entry must be a lowercase domain token matching `^[a-z0-9]+(?:[._-][a-z0-9]+)*$`; the reserved value `swarm-plugin` and any `runner:*` prefix are rejected.
+
 **Skill routing file format** (`.opencode/skill-routing.yaml`):
 
 ```yaml
@@ -695,9 +696,9 @@ Every candidate passes a 3-gate pipeline before entering quarantine:
 | Key | Type | Default | Description |
 |-----|------|---------|-------------|
 | `context_budget.enabled` | boolean | `true` | Enable or disable the context budget guard entirely |
-| `context_budget.max_injection_tokens` | number | `4000` | Token budget for swarm-injected context per turn. This is NOT the model's context window — it's the swarm plugin's own contribution |
-| `context_budget.warn_threshold` | number | `0.7` | Ratio (0.0-1.0) of `max_injection_tokens` that triggers a warning advisory |
-| `context_budget.critical_threshold` | number | `0.9` | Ratio (0.0-1.0) of `max_injection_tokens` that triggers a critical alert with handoff recommendation |
+| `context_budget.max_injection_tokens` | number | `4000` | Separate per-turn cap on system-enhancer injection. It is NOT the guard's budget — the guard measures total conversation tokens against `model_limits`, not this value |
+| `context_budget.warn_threshold` | number | `0.7` | Ratio (0.0-1.0) of the model context window (`model_limits`, default `128000`) at which total conversation tokens trigger a warning advisory |
+| `context_budget.critical_threshold` | number | `0.9` | Ratio (0.0-1.0) of the model context window (`model_limits`, default `128000`) at which total conversation tokens trigger a critical alert with handoff recommendation |
 | `context_budget.enforce` | boolean | `true` | When true, enforces budget limits and may trigger handoffs |
 | `context_budget.prune_target` | number | `0.7` | Ratio (0.0-1.0) of context to preserve when pruning occurs |
 | `context_budget.preserve_last_n_turns` | number | `4` | Number of recent turns to preserve when pruning |
@@ -847,7 +848,7 @@ Swarm uses file locking to protect shared state files from concurrent write corr
 
 ### Locking Implementation
 
-- **Library**: `proper-lockfile` with `retries: 0` (fail-fast — no polling retries)
+- **Library**: `proper-lockfile` with automatic retries (F-09): 5 retries with exponential backoff (10ms→500ms, factor 2); a held lock is treated as stale after `LOCK_TIMEOUT_MS` (5 minutes)
 - **Scope**: Each tool acquires an exclusive lock on the target file before writing
 - **Agents**: Lock is tagged with the current agent name and task context for diagnostics
 

package/dist/agents/template.d.ts

@@ -1,22 +1,29 @@
 /**
- * Agent prompt template renderer.
+ * Agent prompt template helpers.
  *
- * Replaces `{{KEY}}` placeholders in agent prompt strings with values from
- * a `ProjectContext` resolved at session-init time. Strict by design:
- * unknown placeholders raise (caught at build time by
- * `tests/unit/agents/template-substitution.test.ts`), so a typo never
- * leaks to the model.
+ * `{{KEY}}` placeholders in agent prompt strings are substituted with values
+ * from a `ProjectContext` resolved at session-init time by the hand-rolled
+ * `.replace()` chains in `src/agents/index.ts` (Chain B) and
+ * `src/agents/architect.ts` (Chain A). `assertNoUnresolvedPlaceholders` is the
+ * post-substitution safety net that guarantees a typo never leaks a raw
+ * `{{KEY}}` to the model — it runs over each agent's FINAL prompt inside the
+ * `createSwarmAgents` loop (`src/agents/index.ts`), after every substitution
+ * chain has run, before the agents are returned.
  *
- * Phase 4b of language-agnostic plugin work. Pinned call site is
- * `src/index.ts:initializeOpenCodeSwarm` immediately before
- * `getAgentConfigs(...)` — see the `withTimeout(2000ms)` wrapping there
- * to honor invariant 1 (plugin init bounded + fail-open).
+ * Invariant 1 (fail-open): the assertion itself throws (unit tests depend on
+ * it), but its production call site in `createSwarmAgents` wraps it in a
+ * try/catch that downgrades a leftover placeholder to a deferred warning and
+ * registers the agent anyway — a user-authored custom prompt may legitimately
+ * contain a literal `{{UPPER_KEY}}` in prose, and plugin init must never abort.
+ * Built-in-prompt regressions stay caught in CI by a test that drains the
+ * warning buffer after `getAgentConfigs()` over the default prompts.
+ *
+ * Phase 4b of language-agnostic plugin work.
  */
 /**
  * Variables available for substitution into agent prompts. Every prompt's
- * `{{KEY}}` placeholders must be a key of this interface; the renderer
- * rejects unknown placeholders. New variables go here AND in
- * `buildProjectContext` in `src/index.ts`.
+ * `{{KEY}}` placeholders must be a key of this interface. New variables go
+ * here AND in `buildProjectContext` in `src/index.ts`.
  */
 export interface ProjectContext {
     PROJECT_LANGUAGE: string;
@@ -68,15 +75,28 @@ export declare function emptyProjectContext(): ProjectContext;
  */
 export declare function escapeForTemplate(s: string): string;
 /**
- * Render `prompt` with `vars`. Replaces every `{{KEY}}` whose KEY is a
- * documented `ProjectContext` field. Unknown placeholders raise.
+ * Post-substitution safety net for the hand-rolled `.replace()` chains that
+ * assemble each agent's FINAL system prompt (Chain A in
+ * `src/agents/architect.ts` and Chain B in `src/agents/index.ts`). Neither
+ * chain validated for leftover placeholders, so a renamed, mistyped, or
+ * newly-added `{{KEY}}` could leak raw template text straight to the model
+ * with no error. This asserts — over the fully-substituted prompt — that no
+ * `{{KEY}}` survived, applied at the reachable production call site (see
+ * `src/agents/index.ts`, immediately before `getAgentConfigs`).
+ *
+ * The scan uses the character class `[A-Z_]+`. It MUST NOT be broadened to
+ * `[^}]+`, `.*?`, or similar: `src/agents/architect.ts` contains a literal
+ * `` `{{...}}` `` in instructional prose (teaching the architect what an
+ * unresolved field looks like). The dots in `...` are not in `[A-Z_]`, so this
+ * guard never matches that prose. A broader class WOULD match it and
+ * false-throw during agent init. The trailing `g` flag only enumerates every
+ * offender for the error message; it does not widen what counts as a
+ * placeholder.
  *
- * The renderer is intentionally simple — single-pass, no nesting, no
- * conditionals, no loops. Agent prompts that need conditional sections
- * should pre-compute a string variable in `buildProjectContext` and
- * substitute it as a single placeholder.
+ * @throws if any `{{KEY}}` placeholder remains, naming the offending key(s)
+ *   and the agent whose prompt they were found in.
  */
-export declare function renderPrompt(prompt: string, vars: ProjectContext): string;
+export declare function assertNoUnresolvedPlaceholders(prompt: string, agentName: string): void;
 /**
  * Convert an array of constraint strings into a bulleted block ready for
  * inclusion in an agent prompt via `{{CODER_CONSTRAINTS}}` etc.

package/dist/cli/{curator-llm-factory-jrnqg90s.js → curator-llm-factory-ez48eq02.js}

@@ -1,20 +1,20 @@
 // @bun
 import {
   createCuratorLLMDelegate
-} from "./index-5ac7rv03.js";
+} from "./index-cs5765s2.js";
 import"./index-yfedche9.js";
 import"./index-c8s9a3zh.js";
 import"./index-9bsmzfk3.js";
 import"./index-a45jq4b7.js";
-import"./index-f6y341yk.js";
+import"./index-yrrqs4b0.js";
 import"./index-scww5b77.js";
 import"./index-ga4ta3cr.js";
 import"./index-80jkseqw.js";
-import"./index-vjvdvjd3.js";
-import"./index-qrnhvhmg.js";
-import"./index-7stsmndb.js";
-import"./index-kjbfry6m.js";
-import"./index-91qesget.js";
+import"./index-x12hvfpf.js";
+import"./index-n18yy0z1.js";
+import"./index-6f6y2rbp.js";
+import"./index-8f2270hc.js";
+import"./index-rk6qhyng.js";
 import"./index-bfwt3abw.js";
 import"./index-g40r5d08.js";
 import"./index-vxv732ex.js";

package/dist/cli/{curator-vk9ec2kf.js → curator-qj970412.js}

@@ -14,20 +14,20 @@ import {
   runCuratorInit,
   runCuratorPhase,
   writeCuratorSummary
-} from "./index-5ac7rv03.js";
+} from "./index-cs5765s2.js";
 import"./index-yfedche9.js";
 import"./index-c8s9a3zh.js";
 import"./index-9bsmzfk3.js";
 import"./index-a45jq4b7.js";
-import"./index-f6y341yk.js";
+import"./index-yrrqs4b0.js";
 import"./index-scww5b77.js";
 import"./index-ga4ta3cr.js";
 import"./index-80jkseqw.js";
-import"./index-vjvdvjd3.js";
-import"./index-qrnhvhmg.js";
-import"./index-7stsmndb.js";
-import"./index-kjbfry6m.js";
-import"./index-91qesget.js";
+import"./index-x12hvfpf.js";
+import"./index-n18yy0z1.js";
+import"./index-6f6y2rbp.js";
+import"./index-8f2270hc.js";
+import"./index-rk6qhyng.js";
 import"./index-bfwt3abw.js";
 import"./index-g40r5d08.js";
 import"./index-vxv732ex.js";

package/dist/cli/{evidence-summary-service-8g594znj.js → evidence-summary-service-fehwj116.js}

@@ -6,7 +6,7 @@ import {
   loadPlanJsonOnly,
   mergeDurableGateEntriesFromEvidence,
   readDurableGateEvidence
-} from "./index-f6y341yk.js";
+} from "./index-yrrqs4b0.js";
 import"./index-scww5b77.js";
 import"./index-q1exe2b3.js";
 import"./index-wh8949ef.js";

package/dist/cli/{guardrail-explain-vf89cv01.js → guardrail-explain-x2vaxp6s.js}

@@ -1,21 +1,21 @@
 // @bun
 import {
   handleGuardrailExplain
-} from "./index-tmcr6svp.js";
-import"./index-5ac7rv03.js";
+} from "./index-ry8nwsq6.js";
+import"./index-cs5765s2.js";
 import"./index-yfedche9.js";
 import"./index-c8s9a3zh.js";
 import"./index-9bsmzfk3.js";
 import"./index-a45jq4b7.js";
-import"./index-f6y341yk.js";
+import"./index-yrrqs4b0.js";
 import"./index-scww5b77.js";
 import"./index-ga4ta3cr.js";
 import"./index-80jkseqw.js";
-import"./index-vjvdvjd3.js";
-import"./index-qrnhvhmg.js";
-import"./index-7stsmndb.js";
-import"./index-kjbfry6m.js";
-import"./index-91qesget.js";
+import"./index-x12hvfpf.js";
+import"./index-n18yy0z1.js";
+import"./index-6f6y2rbp.js";
+import"./index-8f2270hc.js";
+import"./index-rk6qhyng.js";
 import"./index-bfwt3abw.js";
 import"./index-g40r5d08.js";
 import"./index-vxv732ex.js";

package/dist/cli/{hive-promoter-0b4ny2mp.js → hive-promoter-ysk5edzw.js}

@@ -6,20 +6,20 @@ import {
   isHiveEligible,
   promoteFromSwarm,
   promoteToHive
-} from "./index-5ac7rv03.js";
+} from "./index-cs5765s2.js";
 import"./index-yfedche9.js";
 import"./index-c8s9a3zh.js";
 import"./index-9bsmzfk3.js";
 import"./index-a45jq4b7.js";
-import"./index-f6y341yk.js";
+import"./index-yrrqs4b0.js";
 import"./index-scww5b77.js";
 import"./index-ga4ta3cr.js";
 import"./index-80jkseqw.js";
-import"./index-vjvdvjd3.js";
-import"./index-qrnhvhmg.js";
-import"./index-7stsmndb.js";
-import"./index-kjbfry6m.js";
-import"./index-91qesget.js";
+import"./index-x12hvfpf.js";
+import"./index-n18yy0z1.js";
+import"./index-6f6y2rbp.js";
+import"./index-8f2270hc.js";
+import"./index-rk6qhyng.js";
 import"./index-bfwt3abw.js";
 import"./index-g40r5d08.js";
 import"./index-vxv732ex.js";

package/dist/cli/{index-7stsmndb.js → index-6f6y2rbp.js}

@@ -4,7 +4,7 @@ import {
   countEntryViolationsInWindow,
   readKnowledgeEvents,
   recordKnowledgeEvent
-} from "./index-kjbfry6m.js";
+} from "./index-8f2270hc.js";
 import {
   jaccardBigram,
   readKnowledge,
@@ -12,7 +12,7 @@ import {
   resolveSwarmKnowledgePath,
   transactKnowledge,
   wordBigrams
-} from "./index-91qesget.js";
+} from "./index-rk6qhyng.js";
 import {
   isActiveStatus
 } from "./index-bfwt3abw.js";
@@ -188,7 +188,7 @@ async function maybeQuarantineOnContradiction(directory, entryId, threshold, win
         alreadyInactive: true
       };
     }
-    const { quarantineEntry } = await import("./knowledge-validator-2knz0d2t.js");
+    const { quarantineEntry } = await import("./knowledge-validator-t9sym59g.js");
     await quarantineEntry(directory, entryId, `repeat_contradiction: ${count} contradicted events in ${windowDays}d`, "auto");
     return { quarantined: true, entryId, contradictionsInWindow: count };
   } catch {

package/dist/cli/{index-kjbfry6m.js → index-8f2270hc.js}

@@ -594,7 +594,7 @@ async function applyKnowledgeVerdictFeedback(directory, options) {
       deltas.push({ id, delta });
     }
     if (deltas.length > 0) {
-      const { bumpKnowledgeConfidenceBatch } = await import("./knowledge-store-ksa1dr2z.js");
+      const { bumpKnowledgeConfidenceBatch } = await import("./knowledge-store-9babt8rd.js");
       await bumpKnowledgeConfidenceBatch(directory, deltas, options?.floorOptions);
     }
     return {