SECURITY REVIEW
Not yet assessed
Review the original instructions and requested permissions before installing.
No security review is available for this catalog entry yet.
Agent Development · Documentation / AGENT SKILL
grill-session-record
0
Save completed grill sessions as raw agent-question and user-answer transcripts.
Grill sub-skill. Persist only completed grill sessions as final raw agent-question/user-answer transcripts. In completed preflight runs, write .agent-runs/<id>/grill.md; otherwise fall back to harness/grill/sessions/<timestamp>.md. Activate only when ending a grill session.
Understand the trade-offs.
SKILL QUALITY
Not yet assessed
How clearly the skill guides your agent, how complete its workflow is, and how you can check the outcome.
No quality assessment is available for this catalog entry yet.
The full skill.
Original instructions from the publisher’s SKILL.md
# Grill: Session Record Grill decisions are not durable in conversation memory. Persist the raw exchange only when the grill is ending. ## Hard timing rule Only write a grill transcript after the grill session is complete and the agent is about to end the grill. Do not create, pre-fill, append to, or update any grill transcript while a grill is still active. Do not write research notes, planning notes, recommendations, assistant answers, summaries, decisions, or open questions into the transcript. The transcript is only the completed raw Q&A. ## Primary path during preflight When ending a completed preflight grill for the current numeric run id, write: ```text .agent-runs/<id>/grill.md ``` This is the Step 1 handoff artifact for `intent.md` and `harness.md`. Do not allocate a numeric run id in this skill. Do not write `.agent-runs/<id>/grill.md` for prototype-only work, ordinary design discussion, or any non-preflight grill. ## Fallback path outside preflight When ending a completed non-preflight grill with no current numeric preflight run id, write: ```text harness/grill/sessions/<YYYYMMDD-HHMMSS>.md ``` ## Format Keep the transcript pure Q&A: ```md # Grill <id-or-timestamp> ## Q1: <question text> A: <user answer> ## Q2: <question text> A: <user answer> ``` Do not summarize, interpret, annotate, add decisions, or add open-branch sections. The user's words are the source; interpretation belongs in `intent.md`, `harness.md`, or later grill sessions. `Q` is always the agent's question to the user. `A` is always the user's answer to that question. Never record user questions to the agent as `Q` entries, and never record assistant explanations as `A` entries.