Records a summary of development session progress, verification, and improvements.

Install

mkdir -p .claude/skills/session-report && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/12497" && unzip -o skill.zip -d .claude/skills/session-report && rm skill.zip

Installs to .claude/skills/session-report

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Capture delivery evidence, commands, changed files, blockers, and standards feedback after a work session.
106 charsno explicit “when” trigger
Beginner

Key capabilities

  • →Collect session facts
  • →Capture evidence of implementation and verification
  • →Record requirements drift
  • →Route to next workflow based on status
  • →Generate a structured session report

How it works

This skill collects information about a work session, including changes, commands, and verification results, to generate a structured report.

Inputs & outputs

You give it
Session goal, changed files, or commands run during the session
You get back
A markdown session report detailing changes, verification, risks, and next steps

When to use session-report

  • →Document progress
  • →Create session report
  • →Handoff work

About this skill

Session Report Skill

[!IMPORTANT] Capture delivery evidence, commands, changed files, blockers, and standards feedback after a work session.

Optional args: slug=<feature>, ticket=<id/url>, mode=interactive|autonomous|channel, channel=<id>, auto_continue=true|false, profile=business|hybrid|technical.

Instructions

When the user asks to perform this workflow, execute the following steps:

Session Report Workflow

Goal: Preserve a concise artifact of what changed, how it was verified, and what should improve.

Steps

  1. Collect session facts:
    • User goal, ticket/PR/spec links, changed files, commands run, test results, screenshots/logs, environments.
    • Task-record core (same contract for SDD and non-SDD delivery): stable item/action ID, canonical source reference plus revision, owner, priority, dependencies, criterion references (formal IDs when governed, the approved record's observable criteria otherwise), verification lane, approval reference, and any unresolved blocker with owner and required input. Keep requirement text in its existing authority — the ticket, docs/prd/prd-<slug>.md, docs/srs/srs-<slug>.md, or docs/srs/srs-task-list-<slug>.md — and link it instead of copying or re-minting it.
    • Distinguish checkout absence from absent capability; search authorized refs/worktrees before concluding absence.
    • MCP/tool calls used when available; otherwise local command summaries and artifact paths.
    • Load common-operator-profile; carry the inherited operator_profile, delivery_mode, and work-item identity from the Handoff Payload without re-inferring them.
  2. Capture evidence:
    • Implementation summary tied to ACs or tasks.
    • Verification commands and PASS/FAIL/BLOCKED/NOT RUN status; record warnings, skips, and NOT RUN separately from PASS. Original RED without prior observed run stays NOT RUN.
    • Remaining risks, assumptions, and follow-up owners.
  3. Record requirements drift and authority state:
    • BRD-lite, PRD, and SRS/FRS updates completed or missing; traceability gaps found during implementation or verification.
    • Keep technical verification, business acceptance, and release as separate referenced decisions; an absent reference is not assessed, never success, and no lifecycle state is inferred from timestamps or filenames.
    • SDD delivery links sdd_progress.py status --state <PATH> --revision <REV> --dirty-identity <ID> --format json|markdown as the generated current-state view; checkpoint completion is not verification, acceptance, or release. Non-SDD delivery records the same work-item and decision references here instead of requiring SDD state.
  4. Offer local preparation:
    • Substantive main-session checkpoint with verified readiness (no associated work still active): offer local retrospective preparation once when the host supports it; otherwise state the manual instructions from common/common-session-retrospective references/local-preparation.md. Trivial session: no prompt.
    • Later keeps passive access; Skip suppresses that checkpoint; only meaningful new work plus a new checkpoint permits another offer. Unreliable readiness stays manual-only.
    • Treat this report, a completion signal and any prepared proposal as claims and candidates, never acceptance or approval.
    • No skill, memory, registry or AGENTS_LEARNING.md edit follows; do not read full reports back into model context.
  5. Persist and route:
    • Review private paths/links and one-way manifests; keep publication separate from activation.
    • Write the run record to artifacts/runs/[slug]/[compactISO]-session-report.json when file writes are allowed.
    • Remaining code work -> implement-feature or dev-fix.
    • Missing evidence -> verify-work or traceability-audit.
    • Standards update -> retro-learn.
    • Release communication -> publish-notes.

Runtime Contract

  • Use at the end of a work session to preserve evidence and follow-ups before context is lost.
  • Load common-decision-discipline and apply its §8 continuation contract at entry and every handoff: the settled result selects the successor and specialized gates dominate.
  • Required inputs: session goal plus changed files or commands run during the session.
  • Return BLOCKED only when no session evidence (changes, commands, or verification) exists to report.

Handoff Payload

  • slug, operator_profile, delivery_mode, work-item reference plus revision, approval reference, goal, changes, verification results, evidence-lane refs, requirement-drift status, risks, follow-ups, next workflow.

Blocking Questions

  • Ask max 3 at a time with a recommended default and 2-3 options.

Output Template

# Session Report

## Goal

## Changes

## Verification

| Command/Check | Result | Evidence |
| --- | --- | --- |
| [check] | [result] | [evidence] |

## Work Item And Evidence
- Mode: brief | standard | extended; work item + revision: [ref @ rev]; owner/approval: [owner / ref]
- Criteria: [refs]; verification lane: [lane]; unresolved blocker: [blocker — owner — input]
- Lanes: technical [ref], acceptance [not assessed], release [not assessed]; SDD status view: [projection ref or n/a]

## BRD/PRD/SRS Update Status

## Risks And Follow-Ups

## Skill Feedback Candidates

## Outcome Report
{schema_version: 1, run_id: "[run-id]", slug: "[slug]", workflow: session-report, feature_status: implemented, started_at: "[timestamp]", completed_at: "[timestamp]", requirement_trace: {brd_objectives: [], requirements: [], acceptance_criteria: [], srs: []}, completed_evidence: [], missing_evidence: [], decision_needed: [], recommended_next_workflow: retro-learn, cost: {source: unavailable}, agent: {identity: "[agent-identity]", model: "[model]"}}

## Next Workflow

## Cost Report
Call `get_session_cost(workflow="session-report")` before final handoff.

When not to use it

  • →When no session evidence exists to report
  • →When the goal is not to preserve evidence and follow-ups

Limitations

  • →Requires session goal plus changed files or commands run
  • →Returns BLOCKED if no session evidence exists
  • →Relies on `get_session_cost` for cost reporting

How it compares

This workflow provides a standardized template and process for documenting work sessions, unlike ad-hoc note-taking.

Compared to similar skills

session-report side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
session-report (this skill)05moNo flagsBeginner
spec-kit-workflow1111moNo flagsIntermediate
product-manager-toolkit3211moReviewBeginner
f2s-req-clarify05moNo flagsIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry