fix(claws): point per-mission workspace at /mission/repo + tool-inventory preamble

Two stacked issues after risk_profile was fixed:

1. Claws had file_edit + 46 other tools available, but the templates
   trained the agents to expect file_read/file_write (older ZeroClaw
   tool names). Result: agent output kept saying "I only have file_read"
   and dumped implementations into the context window as text.

2. Even with file_edit, the sandbox pointed at
   /zeroclaw-data/.zeroclaw/agents/<alias>/workspace/ — NOT
   /mission/repo where the checked-out mission repo actually lives.
   unrestricted_filesystem=false blocked agents from reaching it.

Fixes:
- provision_claw now takes workspace_path. mission_orchestrator passes
  /mission/repo — pins the per-claw workspace via
  agents.<alias>.workspace.path to the bind-mount path so file_edit /
  content_search / glob_search operate on the mission's git checkout.
- phase_task_text prepends an explicit tool inventory (file_edit,
  content_search, glob_search, git_operations, git_forge, ...) plus a
  WORKSPACE line pinned at /mission/repo. Each phase directive is
  rewritten to reference file_edit / git_operations explicitly and to
  call out "do NOT paste code in your reply expecting the platform to
  save it."

crates/cm-api/src/mission_orchestrator.rs

@@ -333,8 +333,18 @@ async fn mint_team_from_template(
         // etc.). Passing "toolfree" — the old default — left every
         // agent with zero tools regardless of what its prompt asked for.
         if let Some(p) = provisioner {
+            // /mission/repo is the bind-mount path inside the per-mission
+            // runtime container (see mission_runtime::ensure_container).
+            // Pinning workspace.path there lets file_edit / glob_search /
+            // content_search actually operate on the mission's checked-out
+            // repo instead of the empty per-agent sandbox.
             if let Err(e) = p
-                .provision_claw(claw_id, default_model, &template.template.risk_profile)
+                .provision_claw(
+                    claw_id,
+                    default_model,
+                    &template.template.risk_profile,
+                    Some("/mission/repo"),
+                )
                 .await
             {
                 eprintln!(

crates/cm-api/src/phase_runner.rs

@@ -265,37 +265,65 @@ async fn launch_phase(
 
 fn phase_task_text(kind: &str, title: &str, description: Option<&str>) -> String {
     let base = description.unwrap_or("").trim();
+    // The prior template-derived system prompts trained agents to look
+    // for `file_read`/`file_write` — tools that no longer exist under
+    // ZeroClaw v0.8+. The current toolset uses `file_edit` (create /
+    // overwrite / patch) plus `content_search`/`glob_search`. Injecting
+    // the real tool inventory + concrete workspace path stops the agent
+    // from hallucinating "I only have file_read" and dumping the entire
+    // implementation into the context window instead of onto disk.
+    let tool_preamble = "\
+        TOOLS AVAILABLE (use these exact names — do NOT assume older tool names like file_read / file_write / bash exist):\n\
+        - file_edit         — create, overwrite, or patch files in your workspace\n\
+        - content_search    — grep across your workspace (regex on file contents)\n\
+        - glob_search       — find files by path glob\n\
+        - git_operations    — git status / add / commit / diff / log\n\
+        - git_forge         — Gitea PR / branch / issue operations\n\
+        - web_search_tool / web_fetch — external references (research-tier profiles only)\n\
+        - spawn_subagent    — hand off a subtask to another claw\n\
+        - delegate          — call a peer role by name\n\
+        - memory_store / memory_recall — durable per-agent notes\n\
+        \n\
+        WORKSPACE: Your working directory is /mission/repo. That path is the\n\
+        mission's git checkout. All file_edit / content_search / glob_search\n\
+        operations resolve there. To read a file: file_edit with mode='read'\n\
+        or content_search first, then file_edit to patch. Write your outputs\n\
+        as REAL files with file_edit — do NOT paste code blocks in your reply\n\
+        expecting the platform to save them; nothing else writes files for you.\n";
     let directive = match kind {
         "research" => {
             "Your team is running the RESEARCH phase of this mission. \
              Investigate the topic, gather sources, and produce a \
              sectioned Markdown brief the coding phase can implement \
-             directly. Save findings to the Obsidian vault or the \
+             directly. Save findings under /mission/repo/research/ \
-             mission's artifact directory. Emit INT-XX task markers \
+             using file_edit — one Markdown file per topic. Emit INT-XX \
-             for concrete follow-ups."
+             task markers in the last file for concrete follow-ups."
         }
         "coding" => {
             "Your team is running the CODING phase of this mission. \
              Implement the mission's acceptance criteria against the \
-             checked-out repo. Follow the workspace-repo-commit-protocol: \
+             /mission/repo checkout using file_edit for every source \
-             small focused commits with test coverage. Emit COMPLETED: <INT-id> \
+             file, then git_operations to commit small focused changes \
-             markers as you close research-produced tasks."
+             with test coverage. Emit COMPLETED: <INT-id> markers as you \
+             close research-produced tasks. Do NOT respond with source \
+             code in text — write it as files."
         }
         "benchmark" => {
             "Your team is running the BENCHMARK phase of this mission. \
-             Author or extend benchmarks that measure the target change. \
+             Author or extend benchmarks under /mission/repo/benches or \
-             Baseline the pre-change performance, apply the change (or \
+             the crate's bench harness using file_edit. Baseline the \
-             use the mission's committed diff), then measure after."
+             pre-change performance, apply the change (or use the \
+             mission's committed diff), then measure after."
         }
         "security_scan" => {
             "Your team is running the SECURITY SCAN phase of this mission. \
-             Run cargo-audit, gitleaks, trivy, and semgrep against the \
+             Run cargo-audit, gitleaks, trivy, and semgrep against \
-             checked-out repo. Triage findings, file INT-XX task markers \
+             /mission/repo. Triage findings, file INT-XX task markers \
              for remediation, propose patches for the coding phase."
         }
         _ => "Execute this mission phase according to the mission brief.",
     };
-    format!("MISSION: {title}\n\n{directive}\n\nBRIEF:\n{base}")
+    format!("MISSION: {title}\n\n{tool_preamble}\n{directive}\n\nBRIEF:\n{base}")
 }
 
 /// Close phases whose topology_runs are all terminal.

crates/cm-api/src/routes/teams.rs

@@ -130,8 +130,11 @@ pub(crate) async fn build_team_with_lifecycle(
         .await?;
         let claw_id = agent.id.as_uuid();
         let risk = RuntimeProvisioner::default_risk_profile_for_role(&m.role);
+        // No workspace override on the ad-hoc team-wizard path — those
+        // teams aren't mission-bound so they use the default per-agent
+        // workspace under <install>/agents/<alias>/workspace/.
         provisioner
-            .provision_claw(claw_id, &m.model, risk)
+            .provision_claw(claw_id, &m.model, risk, None)
             .await
             .map_err(|e| {
                 eprintln!("teams: provision claw {claw_id} failed: {e}");

crates/cm-api/src/runtime_provision.rs

@@ -136,12 +136,22 @@ impl RuntimeProvisioner {
     /// `risk_profile` (from the team template — controls which tools this
     /// agent gets: `toolfree` = nothing, `research_readonly` = file_read
     /// only, `coding_readwrite` = file_read + file_write + shell, etc.),
-    /// and the `clawmates_door` MCP bundle. Idempotent on the create step.
+    /// and the `clawmates_door` MCP bundle.
+    ///
+    /// `workspace_path`, when Some, pins the agent's per-agent workspace
+    /// dir via `[agents.<alias>.workspace.path]`. In per-mission runtime
+    /// containers this is `/mission/repo` so `file_edit` / `content_search`
+    /// / `glob_search` operate on the mission's checked-out repo instead
+    /// of the default `<install>/agents/<alias>/workspace/` sandbox
+    /// (which the agent can't populate with the mission's source files).
+    ///
+    /// Idempotent on the create step.
     pub async fn provision_claw(
         &self,
         claw_id: Uuid,
         model: &str,
         risk_profile: &str,
+        workspace_path: Option<&str>,
     ) -> Result<String, String> {
         let alias = claw_alias(claw_id);
         let model_alias = provider_alias_for(model);
@@ -179,6 +189,13 @@ impl RuntimeProvisioner {
             serde_json::json!(["clawmates_door"]),
         )
         .await?;
+        if let Some(path) = workspace_path {
+            self.set_prop(
+                &format!("agents.{alias}.workspace.path"),
+                serde_json::json!(path),
+            )
+            .await?;
+        }
 
         Ok(alias)
     }