MCP HubMCP Hub
Retour aux compétences

read-continue-here

pjt222
Mis à jour 2 days ago
8 vues
17
2
17
Voir sur GitHub
Designaiautomation

À propos

Cette compétence permet à Claude de détecter automatiquement et de reprendre le travail d'une session antérieure en lisant un fichier `CONTINUE_HERE.md` au démarrage. Elle analyse la passation structurée, confirme le plan de reprise avec l'utilisateur et nettoie après consommation. Utilisez-la lors de l'amorçage après des interruptions ou pour configurer la détection automatique de continuation via les hooks SessionStart.

Installation rapide

Claude Code

Recommandé
Principal
npx skills add pjt222/agent-almanac -a claude-code
Commande PluginAlternatif
/plugin add https://github.com/pjt222/agent-almanac
Git CloneAlternatif
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/read-continue-here

Copiez et collez cette commande dans Claude Code pour installer cette compétence

Documentation

Read Continue Here

Read a structured continuation file and resume work from where the prior session left off.

When to Use

  • Starting a new session and CONTINUE_HERE.md exists in the project root
  • After a SessionStart hook injects continuation context
  • Bootstrapping identity and detecting prior session artifacts
  • Setting up automatic continuation detection for a project (one-time infrastructure)

Inputs

  • Required: A project directory (defaults to current working directory)
  • Optional: Whether to configure infrastructure (SessionStart hook + CLAUDE.md instruction)
  • Optional: Whether to delete the file after consumption (default: yes)

Procedure

Step 1: Detect and Read the Continuation File

Check for CONTINUE_HERE.md in the project root:

ls -la CONTINUE_HERE.md 2>/dev/null

If absent, exit gracefully — there is nothing to continue from.

If present, read the file contents. Parse the 5 sections: Objective, Completed, In Progress, Next Steps, Context. Extract the timestamp and branch from the header line.

Got: The file is read and its sections are parsed into a clear mental model of the prior session's state.

If fail: If the file exists but is malformed (missing sections, empty), treat it as a partial signal — extract whatever is present and note what is missing to the user.

Step 2: Assess Freshness

Compare the file's timestamp against the current time:

# File modification time
stat -c '%Y' CONTINUE_HERE.md 2>/dev/null || stat -f '%m' CONTINUE_HERE.md
# Current time
date +%s

Classify freshness:

  • Fresh (< 24 hours, same branch): safe to act on directly
  • Stale (> 24 hours or different branch): flag to user before proceeding
  • Superseded (new commits exist after the handoff timestamp): someone worked on the project since the handoff

Check branch alignment:

git branch --show-current
git log --oneline --since="$(stat -c '%Y' CONTINUE_HERE.md | xargs -I{} date -d @{} --iso-8601=seconds)" 2>/dev/null

Got: A freshness assessment with classification (fresh, stale, or superseded) and supporting evidence.

If fail: If not in a git repo, skip branch and commit checks. Rely on the timestamp in the file header alone.

Step 3: Summarize and Confirm Resumption

Present the continuation state to the user concisely:

  • "Prior session objective: [Objective]"
  • "Completed: [summary]"
  • "In progress: [summary]"
  • "Proposed next action: [Next Steps item 1]"

If freshness is "stale" or "superseded", present the evidence and ask whether to proceed with the handoff or start fresh.

If any Next Steps items are tagged **[USER]**, surface those explicitly — they require user decisions before work can proceed.

Got: The user confirms the resumption plan, possibly with adjustments. The agent has a clear mandate for what to do next.

If fail: If the user says "start fresh" or "ignore that file", acknowledge and proceed without the continuation context. Offer to delete the file to prevent future confusion.

Step 4: Act on the Handoff

Begin working from Next Steps item 1 (or wherever the user directed):

  • Reference In Progress items to understand partial state
  • Use the Context section to avoid retrying failed approaches
  • Treat Completed items as done — do not re-verify unless the user asks

Got: The agent is productively working on the right task, informed by the continuation file.

If fail: If the Next Steps are ambiguous or the In Progress state is unclear, ask the user for clarification rather than guessing.

Step 5: Clean Up

After the handoff is consumed and work is underway, delete CONTINUE_HERE.md:

rm CONTINUE_HERE.md

Stale continuation files cause confusion in future sessions.

Got: The file is removed. The project root is clean.

If fail: If the user wants to keep the file (e.g., as a reference during the session), leave it but note that it should be deleted before session end to prevent the next session from re-consuming it.

Step 6: Configure SessionStart Hook (Optional)

If not already configured, set up automatic reading of CONTINUE_HERE.md on session start.

Create the hook script:

mkdir -p ~/.claude/hooks/continue-here

cat > ~/.claude/hooks/continue-here/read-continuation.sh << 'SCRIPT'
#!/bin/bash
# SessionStart hook: inject CONTINUE_HERE.md into session context
# OS-aware: works on native Linux, WSL, macOS, and Windows (Git Bash/MSYS)
set -uo pipefail

# --- Platform detection ---
detect_platform() {
  case "$(uname -s)" in
    Darwin) echo "mac" ;;
    Linux)
      if grep -qi microsoft /proc/version 2>/dev/null; then
        echo "wsl"
      else
        echo "linux"
      fi ;;
    MINGW*|MSYS*|CYGWIN*) echo "windows" ;;
    *) echo "unknown" ;;
  esac
}
PLATFORM=${PLATFORM:-$(detect_platform)}

CONTINUE_FILE="$PWD/CONTINUE_HERE.md"

if [ ! -f "$CONTINUE_FILE" ]; then
  exit 0
fi

# Strip CRLF (files on NTFS often have Windows line endings)
CONTENT=$(sed 's/\r$//' "$CONTINUE_FILE")

# JSON-escape: prefer jq, fall back to portable awk
if command -v jq >/dev/null 2>&1; then
  ESCAPED=$(printf '%s' "$CONTENT" | jq -Rsa .)
else
  ESCAPED=$(printf '%s' "$CONTENT" | awk '
    BEGIN { ORS=""; print "\"" }
    {
      gsub(/\\/, "\\\\")
      gsub(/"/, "\\\"")
      gsub(/\t/, "\\t")
      if (NR > 1) print "\\n"
      print
    }
    END { print "\"" }
  ')
fi

cat << EOF
{"hookSpecificOutput":{"sessionStartContext":{"additionalContext":$ESCAPED}}}
EOF
SCRIPT

chmod +x ~/.claude/hooks/continue-here/read-continuation.sh

Add to ~/.claude/settings.json in the SessionStart hooks array:

{
  "type": "command",
  "command": "~/.claude/hooks/continue-here/read-continuation.sh",
  "timeout": 5
}

Got: The hook script exists, is executable, and is registered in settings.json. On next session start, if CONTINUE_HERE.md exists, its content is injected into the session context.

If fail: Check that settings.json is valid JSON after editing. Test the hook manually: cd /your/project && ~/.claude/hooks/continue-here/read-continuation.sh. The script falls back to awk if jq is not installed, so jq is recommended but not required.

Step 7: Add CLAUDE.md Instruction (Optional)

Add a brief instruction to the project's CLAUDE.md so Claude understands the file's purpose:

## Session Continuity

If `CONTINUE_HERE.md` exists in the project root, read it at session start. It contains a structured handoff from a prior session: objective, completed work, in-progress state, next steps, and context. Act on it — acknowledge the continuation, summarize prior state, and propose resuming from the Next Steps section. If the file is older than 24 hours, flag this to the user before proceeding. After the handoff is consumed, the file can be deleted.

Got: CLAUDE.md contains the instruction. Future sessions will read and act on CONTINUE_HERE.md even if the SessionStart hook is not configured.

If fail: If CLAUDE.md does not exist, create it with just this section. If the file is too long, add the instruction near the top where it will not be truncated.

Validation

  • CONTINUE_HERE.md was detected (or absence was handled gracefully)
  • Freshness was assessed (timestamp, branch, post-handoff commits)
  • Resumption plan was presented to and confirmed by the user
  • Work began from the correct Next Steps item
  • The file was cleaned up after consumption
  • (Optional) SessionStart hook script exists and is executable
  • (Optional) CLAUDE.md contains the session continuity instruction

Pitfalls

  • Acting without confirming: Always present the resumption plan to the user. They may have changed their mind about what to work on, even if the file is fresh.
  • Trusting stale files blindly: A continuation file older than 24 hours or from a different branch is a suggestion, not a mandate. Always check freshness.
  • Ignoring the Context section: The most valuable part of the file is often the failed approaches. Skipping this section leads to retrying dead ends.
  • Forgetting to clean up: Leaving CONTINUE_HERE.md after consumption causes confusion in the next session, which will try to act on it again.
  • Treating Completed items as unverified: Unless the user specifically asks, do not re-do completed work. Trust the prior session's assessment.

Related Skills

  • write-continue-here — the complement: writing the continuation file at session end
  • bootstrap-agent-identity — full identity reconstruction that includes continuation detection as one heuristic
  • manage-memory — durable cross-session knowledge (complements this ephemeral handoff)
  • write-claude-md — project instructions where the optional continuity guidance lives

Dépôt GitHub

pjt222/agent-almanac
Chemin: i18n/caveman-lite/skills/read-continue-here
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

Compétences associées

executing-plans

Design

Utilisez la compétence executing-plans lorsque vous disposez d'un plan de mise en œuvre complet à exécuter par lots contrôlés avec des points de contrôle de revue. Elle charge et examine le plan de manière critique, puis exécute les tâches par petits lots (3 tâches par défaut) tout en rapportant la progression entre chaque lot pour une revue par l'architecte. Cela garantit une mise en œuvre systématique avec des points de contrôle de qualité intégrés.

Voir la compétence

requesting-code-review

Design

Cette compétence délègue un sous-agent réviseur de code pour analyser les modifications apportées au code par rapport aux exigences avant de poursuivre. Elle doit être utilisée après avoir terminé des tâches, implémenté des fonctionnalités majeures, ou avant une fusion vers la branche principale. La revue aide à détecter précocement les problèmes en comparant l'implémentation actuelle avec le plan initial.

Voir la compétence

connect-mcp-server

Design

Cette compétence fournit un guide complet permettant aux développeurs de connecter des serveurs MCP à Claude Code via les transports HTTP, stdio ou SSE. Elle couvre l'installation, la configuration, l'authentification et la sécurité pour intégrer des services externes tels que GitHub, Notion et des API personnalisées. Utilisez-la lors de la configuration d'intégrations MCP, de la configuration d'outils externes ou du travail avec le Protocole de Contexte de Modèle de Claude.

Voir la compétence

web-cli-teleport

Design

Cette compétence aide les développeurs à choisir entre les interfaces Web et CLI de Claude Code en fonction de l'analyse des tâches, puis permet une téléportation transparente des sessions entre ces environnements. Elle optimise le flux de travail en gérant l'état et le contexte de la session lors du passage entre le web, la CLI ou le mobile. Utilisez-la pour des projets complexes nécessitant différents outils à diverses étapes.

Voir la compétence