diff --git a/skills/dispatching-parallel-agents/SKILL.md b/skills/dispatching-parallel-agents/SKILL.md
index 75e7e22cef..802448681e 100644
--- a/skills/dispatching-parallel-agents/SKILL.md
+++ b/skills/dispatching-parallel-agents/SKILL.md
@@ -83,6 +83,7 @@ When agents return:
 - Verify fixes don't conflict
 - Run full test suite
 - Integrate all changes
+- Release each agent unless you intend to send it further input — on some harnesses a finished agent holds its slot until closed (see the per-platform tool refs in `../using-superpowers/references/`)
 
 ## Agent Prompt Structure
 
diff --git a/skills/requesting-code-review/SKILL.md b/skills/requesting-code-review/SKILL.md
index 4b8aa605fc..90c6947f73 100644
--- a/skills/requesting-code-review/SKILL.md
+++ b/skills/requesting-code-review/SKILL.md
@@ -44,6 +44,7 @@ Dispatch a `general-purpose` subagent, filling the template at [code-reviewer.md
 - Fix Important issues before proceeding
 - Note Minor issues for later
 - Push back if reviewer is wrong (with reasoning)
+- Release the reviewer unless you intend to send it further input — on some harnesses a finished agent holds its slot until closed (see the per-platform tool refs in `../using-superpowers/references/`)
 
 ## Example
 
diff --git a/skills/subagent-driven-development/SKILL.md b/skills/subagent-driven-development/SKILL.md
index d8ca081570..464f332958 100644
--- a/skills/subagent-driven-development/SKILL.md
+++ b/skills/subagent-driven-development/SKILL.md
@@ -215,6 +215,8 @@ final whole-branch review. When you fill a reviewer template:
   subagent with the complete findings list — not one fixer per finding.
   Per-finding fixers each rebuild context and re-run suites; a real
   session's final-review fix wave cost more than all its tasks combined.
+- Release the final whole-branch reviewer, and any fixer it triggered, once
+  you have acted on its findings and will send them no further input.
 
 ## File Handoffs
 
@@ -257,6 +259,11 @@ a ledger file, not only in todos.
 - When a task's review comes back clean, append one line to the ledger in
   the same message as your other bookkeeping:
   `Task N: complete (commits <base7>..<head7>, review clean)`.
+  That bookkeeping includes releasing the subagents the task used —
+  implementer, task reviewer, and any fix subagents — you will send none
+  of them further input. On some harnesses a finished agent holds its slot
+  until closed (see the per-platform tool refs in
+  `../using-superpowers/references/`).
 - The ledger is your recovery map: the commits it names exist in git even
   when your context no longer remembers creating them. After compaction,
   trust the ledger and `git log` over your own recollection.