job-forge 2.14.4 → 2.14.6

package/.codex/config.toml

@@ -15,11 +15,6 @@ model_reasoning_effort = "high"
 model = "gpt-5.4-nano"
 model_provider = "openai"
 
-[model_providers.openai]
-name = "OpenAI"
-base_url = "https://api.openai.com/v1"
-env_key = "OPENAI_API_KEY"
-
 [mcp_servers.geometra]
 command = "npx"
 args = ["-y", "@geometra/mcp"]

package/.cursor/rules/main.mdc

@@ -304,6 +304,8 @@ Mode routing is specified in the top-level **## Routing** section. Each mode is
 
 ## Offer Verification -- MANDATORY
 
+**Read local artifacts before the network.** If `reports/` already contains this posting URL (or company+role with a full JD in the body), **Read** that report for verification or evaluation instead of WebFetch/Geometra. If `data/pipeline.md` or `jds/` points at frozen JD text (`local:jds/{file}` or pasted blocks), **Read** that first. Reuse JD text already in the same conversation — do not fetch the same URL twice. (The JD extraction section at the top of `modes/auto-pipeline.md` and its "at most once per session" rule are the detailed contract.)
+
 **When Geometra MCP is available** (interactive sessions), ALWAYS use it to verify offers:
 1. `geometra_connect` to the URL (via proxy)
 2. `geometra_page_model` to read structured page content
@@ -317,6 +319,17 @@ Mode routing is specified in the top-level **## Routing** section. Each mode is
 
 The goal is to never waste time on closed offers, but also never silently assume a role is active when verification was incomplete.
 
+### Canonical MCP tools (quick reference)
+
+Pick tools by name directly — reduces unnecessary tool discovery:
+
+| Task | Preferred tools |
+|------|------------------|
+| JD from URL | Greenhouse boards API when the URL matches (see JD extraction in `modes/auto-pipeline.md`) → else `geometra_connect` + `geometra_page_model` → else WebFetch → WebSearch last |
+| Offer still live? | Same as JD when Geometra is available; else WebFetch per above |
+| One apply subagent (single job) | One `geometra_connect` per job URL; reuse `sessionId` through schema + fill; submit via atomic `geometra_run_actions` per `modes/apply.md` [H1]. Do **not** `geometra_disconnect` between `geometra_form_schema` and submit on the same form unless recovery requires it |
+| Chromium pool between orchestrator dispatch rounds | `geometra_list_sessions` + `geometra_disconnect({ closeBrowser: true })` per Hard limit [H3] — orchestrator-only; not a substitute for finishing the in-subagent form flow |
+
 ---
 
 ## OTP Handling via Gmail MCP -- REQUIRED

package/AGENTS.md

@@ -299,6 +299,8 @@ Mode routing is specified in the top-level **## Routing** section. Each mode is
 
 ## Offer Verification -- MANDATORY
 
+**Read local artifacts before the network.** If `reports/` already contains this posting URL (or company+role with a full JD in the body), **Read** that report for verification or evaluation instead of WebFetch/Geometra. If `data/pipeline.md` or `jds/` points at frozen JD text (`local:jds/{file}` or pasted blocks), **Read** that first. Reuse JD text already in the same conversation — do not fetch the same URL twice. (The JD extraction section at the top of `modes/auto-pipeline.md` and its "at most once per session" rule are the detailed contract.)
+
 **When Geometra MCP is available** (interactive sessions), ALWAYS use it to verify offers:
 1. `geometra_connect` to the URL (via proxy)
 2. `geometra_page_model` to read structured page content
@@ -312,6 +314,17 @@ Mode routing is specified in the top-level **## Routing** section. Each mode is
 
 The goal is to never waste time on closed offers, but also never silently assume a role is active when verification was incomplete.
 
+### Canonical MCP tools (quick reference)
+
+Pick tools by name directly — reduces unnecessary tool discovery:
+
+| Task | Preferred tools |
+|------|------------------|
+| JD from URL | Greenhouse boards API when the URL matches (see JD extraction in `modes/auto-pipeline.md`) → else `geometra_connect` + `geometra_page_model` → else WebFetch → WebSearch last |
+| Offer still live? | Same as JD when Geometra is available; else WebFetch per above |
+| One apply subagent (single job) | One `geometra_connect` per job URL; reuse `sessionId` through schema + fill; submit via atomic `geometra_run_actions` per `modes/apply.md` [H1]. Do **not** `geometra_disconnect` between `geometra_form_schema` and submit on the same form unless recovery requires it |
+| Chromium pool between orchestrator dispatch rounds | `geometra_list_sessions` + `geometra_disconnect({ closeBrowser: true })` per Hard limit [H3] — orchestrator-only; not a substitute for finishing the in-subagent form flow |
+
 ---
 
 ## OTP Handling via Gmail MCP -- REQUIRED

package/CLAUDE.md

@@ -299,6 +299,8 @@ Mode routing is specified in the top-level **## Routing** section. Each mode is
 
 ## Offer Verification -- MANDATORY
 
+**Read local artifacts before the network.** If `reports/` already contains this posting URL (or company+role with a full JD in the body), **Read** that report for verification or evaluation instead of WebFetch/Geometra. If `data/pipeline.md` or `jds/` points at frozen JD text (`local:jds/{file}` or pasted blocks), **Read** that first. Reuse JD text already in the same conversation — do not fetch the same URL twice. (The JD extraction section at the top of `modes/auto-pipeline.md` and its "at most once per session" rule are the detailed contract.)
+
 **When Geometra MCP is available** (interactive sessions), ALWAYS use it to verify offers:
 1. `geometra_connect` to the URL (via proxy)
 2. `geometra_page_model` to read structured page content
@@ -312,6 +314,17 @@ Mode routing is specified in the top-level **## Routing** section. Each mode is
 
 The goal is to never waste time on closed offers, but also never silently assume a role is active when verification was incomplete.
 
+### Canonical MCP tools (quick reference)
+
+Pick tools by name directly — reduces unnecessary tool discovery:
+
+| Task | Preferred tools |
+|------|------------------|
+| JD from URL | Greenhouse boards API when the URL matches (see JD extraction in `modes/auto-pipeline.md`) → else `geometra_connect` + `geometra_page_model` → else WebFetch → WebSearch last |
+| Offer still live? | Same as JD when Geometra is available; else WebFetch per above |
+| One apply subagent (single job) | One `geometra_connect` per job URL; reuse `sessionId` through schema + fill; submit via atomic `geometra_run_actions` per `modes/apply.md` [H1]. Do **not** `geometra_disconnect` between `geometra_form_schema` and submit on the same form unless recovery requires it |
+| Chromium pool between orchestrator dispatch rounds | `geometra_list_sessions` + `geometra_disconnect({ closeBrowser: true })` per Hard limit [H3] — orchestrator-only; not a substitute for finishing the in-subagent form flow |
+
 ---
 
 ## OTP Handling via Gmail MCP -- REQUIRED

package/docs/CUSTOMIZATION.md

@@ -90,6 +90,24 @@ JobForge can integrate with external systems via opencode hooks. Example hooks:
 
 Save hooks in `.opencode/settings.json`.
 
+## Application tracker layouts
+
+The default harness uses **day-based** tracker files: `data/applications/YYYY-MM-DD.md` (one markdown table per calendar day).
+
+Some forks use a **single** `data/applications.md` instead. That is fine if you document it in the project root `AGENTS.md` (or equivalent) and keep `npx job-forge merge` / `verify` aligned with whatever `merge-tracker.mjs` and `verify-pipeline.mjs` expect in your fork.
+
+## Transcript observability (iso-trace)
+
+To inspect real agent sessions locally (tool mix, redundant fetches, Geometra churn) without uploading transcripts, use Razroo's [`@razroo/iso-trace`](https://github.com/razroo/iso/tree/main/packages/iso-trace).
+
+This harness also ships npm scripts (from the package root): `npm run trace:list`, `npm run trace:stats`, and `npm run trace:show -- <id>` — they wrap `iso-trace` with sensible defaults when the CLI is on `PATH`.
+
+**Where Claude Code writes JSONL:** `~/.claude/projects/<encoded-cwd>/*.jsonl`.
+
+**CLI (install once per machine):** `npx -y @razroo/iso-trace@latest stats --source "$HOME/.claude/projects/<encoded-dir>/<session>.jsonl"`
+
+**Performance:** `iso-trace list --cwd /path/to/repo` walks all of `~/.claude/projects` before filtering; on large machines prefer `stats --source <one.jsonl>` or the library's `discoverSessions({ roots: ["<one encoded project dir>"] })` (see the iso-trace README).
+
 ## States (templates/states.yml)
 
 The canonical states rarely need changing. Since `templates/` is a symlink into the harness in consumer projects, adding new states means contributing back to `razroo/JobForge` (see [CONTRIBUTING.md](../CONTRIBUTING.md)). If you're working in the harness repo directly (Path B), update:

package/iso/instructions.md

@@ -299,6 +299,8 @@ Mode routing is specified in the top-level **## Routing** section. Each mode is
 
 ## Offer Verification -- MANDATORY
 
+**Read local artifacts before the network.** If `reports/` already contains this posting URL (or company+role with a full JD in the body), **Read** that report for verification or evaluation instead of WebFetch/Geometra. If `data/pipeline.md` or `jds/` points at frozen JD text (`local:jds/{file}` or pasted blocks), **Read** that first. Reuse JD text already in the same conversation — do not fetch the same URL twice. (The JD extraction section at the top of `modes/auto-pipeline.md` and its "at most once per session" rule are the detailed contract.)
+
 **When Geometra MCP is available** (interactive sessions), ALWAYS use it to verify offers:
 1. `geometra_connect` to the URL (via proxy)
 2. `geometra_page_model` to read structured page content
@@ -312,6 +314,17 @@ Mode routing is specified in the top-level **## Routing** section. Each mode is
 
 The goal is to never waste time on closed offers, but also never silently assume a role is active when verification was incomplete.
 
+### Canonical MCP tools (quick reference)
+
+Pick tools by name directly — reduces unnecessary tool discovery:
+
+| Task | Preferred tools |
+|------|------------------|
+| JD from URL | Greenhouse boards API when the URL matches (see JD extraction in `modes/auto-pipeline.md`) → else `geometra_connect` + `geometra_page_model` → else WebFetch → WebSearch last |
+| Offer still live? | Same as JD when Geometra is available; else WebFetch per above |
+| One apply subagent (single job) | One `geometra_connect` per job URL; reuse `sessionId` through schema + fill; submit via atomic `geometra_run_actions` per `modes/apply.md` [H1]. Do **not** `geometra_disconnect` between `geometra_form_schema` and submit on the same form unless recovery requires it |
+| Chromium pool between orchestrator dispatch rounds | `geometra_list_sessions` + `geometra_disconnect({ closeBrowser: true })` per Hard limit [H3] — orchestrator-only; not a substitute for finishing the in-subagent form flow |
+
 ---
 
 ## OTP Handling via Gmail MCP -- REQUIRED

package/modes/apply.md

@@ -36,6 +36,9 @@ Live application assistant. Reads the active application form in Chrome (via Geo
 - [D5] Fetch `geometra_form_schema` at most once per application, right after the initial `geometra_connect`. Operate on labels thereafter.
   why: schema re-fetches return hundreds of nested field IDs and pollute context; labels don't change mid-flow, so the second fetch is just paying for the same payload twice
 
+- [D5b] Reuse the same Geometra `sessionId` for every tool call on that job URL until the atomic submit transaction completes (or recovery sequence finishes). Do not `geometra_disconnect` mid-flow between schema and submit unless [H1] recovery forces it.
+  why: disconnecting between schema and fill invalidates the SPA session and matches real traces where redundant disconnects outnumbered connects; orchestrator pool cleanup [H3]/[H4] is a separate concern between dispatch rounds, not inside one apply subagent
+
 - [D6] Use `fieldLabel` over `fieldId` everywhere it works.
   why: labels are stable across DOM refreshes; IDs are regenerated
 

package/modes/auto-pipeline.md

@@ -21,6 +21,8 @@ Fetch the JD content once. If the input is a **URL** (not pasted JD text), fetch
 
 **If the input is JD text** (not a URL): use it directly, no fetching needed.
 
+**Local artifacts before Step 0 methods:** Grep `reports/` for the URL or stable company+role slug; if a report already embeds the full JD, Read it and skip network fetch entirely. If the pipeline row or `jds/` references `local:jds/{file}`, Read that file first. This stacks with the rule above: one fetch per URL per session, and **zero** if the JD is already on disk.
+
 ## Step 1 — Run Evaluation A-F
 Execute exactly as in the `offer` mode (read `modes/offer.md` for all blocks A-F).
 
@@ -34,7 +36,7 @@ Execute the full `pdf` pipeline (read `modes/pdf.md`).
 
 Generate draft answers for the application form when the final score is >= 3.5. If the final score is >= 3.5 (per Canonical Scoring Model thresholds in `_shared.md`), generate draft answers for the application form:
 
-1. **Extract form questions**: Use Geometra MCP (`geometra_connect` + `geometra_form_schema`) to discover all form fields. If questions cannot be extracted, use the generic questions.
+1. **Extract form questions**: Use Geometra MCP (`geometra_connect` + `geometra_form_schema`) to discover all form fields. **Reuse the same `sessionId` from Step 0** when the apply URL is the same rendered page; only connect again if the prior session ended or the URL changed. If questions cannot be extracted, use the generic questions.
 2. **Generate answers** following the tone guidelines (see below).
 3. **Save in the report** as a `## G) Draft Application Answers` section.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "job-forge",
-  "version": "2.14.4",
+  "version": "2.14.6",
   "description": "AI-powered job search pipeline built on opencode",
   "type": "module",
   "bin": {
@@ -89,7 +89,7 @@
   "devDependencies": {
     "@razroo/iso": "^0.2.5",
     "@razroo/iso-harness": "^0.6.1",
-    "@razroo/iso-route": "^0.5.0",
+    "@razroo/iso-route": "^0.5.2",
     "@razroo/iso-trace": "^0.3.1",
     "@razroo/opencode-model-fallback": "^0.3.1"
   }