diff --git a/codex-rs/core/templates/compact/prompt.md b/codex-rs/core/templates/compact/prompt.md
index 42fae605db8..e78263dfa3c 100644
--- a/codex-rs/core/templates/compact/prompt.md
+++ b/codex-rs/core/templates/compact/prompt.md
@@ -1,9 +1,15 @@
-You are performing a CONTEXT CHECKPOINT COMPACTION. Create a handoff summary for another LLM that will resume the task.
+You are performing a CONTEXT CHECKPOINT COMPACTION. Create a handoff summary so another LLM can continue the single active task (the most recent user request that is still in progress).
 
-Include:
-- Current progress and key decisions made
-- Important context, constraints, or user preferences
-- What remains to be done (clear next steps)
-- Any critical data, examples, or references needed to continue
+Rules:
+- Focus only on the active task. Do not restate or reopen earlier, completed tasks.
+- Do not propose audits or re-checking unless the user explicitly asked.
+- If earlier tasks exist, note them as completed in one short bullet.
 
-Be concise, structured, and focused on helping the next LLM seamlessly continue the work.
+Use this format:
+- Active task:
+- Progress so far:
+- Next steps (active task only):
+- Key context / constraints / preferences (still relevant only):
+- Completed tasks (optional, brief):
+
+Be concise and structured so the next LLM can continue without broad backtracking.
diff --git a/codex-rs/core/templates/compact/summary_prefix.md b/codex-rs/core/templates/compact/summary_prefix.md
index 62a7161b89b..99b5c379e27 100644
--- a/codex-rs/core/templates/compact/summary_prefix.md
+++ b/codex-rs/core/templates/compact/summary_prefix.md
@@ -1 +1 @@
-Another language model started to solve this problem and produced a summary of its thinking process. You also have access to the state of the tools that were used by that language model. Use this to build on the work that has already been done and avoid duplicating work. Here is the summary produced by the other language model, use the information in this summary to assist with your own analysis:
\ No newline at end of file
+You are resuming an in-progress task after context compaction. Another language model produced a handoff summary. Treat the summary as authoritative for what is already done. Continue from the "Next steps" for the active task; do not re-open completed tasks or audit prior work unless the user explicitly asked. You also have access to the state of the tools that were used by that language model. Here is the summary: