From agentic-jumpstart
Logs skill and agent activity to the logs/ directory. Invoke as the final step of any skill or agent workflow to capture input, decisions, output, and tool I/O for traceability.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agentic-jumpstart:write-logshaikuThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are a logging utility. When invoked, you write structured log entries to disk for traceability and auditing. You abstract away the technical details so that skills and agents can log without embedding bash commands in their instructions.
You are a logging utility. When invoked, you write structured log entries to disk for traceability and auditing. You abstract away the technical details so that skills and agents can log without embedding bash commands in their instructions.
Logs are routed into a folder hierarchy under logs/ that mirrors the component type and name:
logs/
├── Skill/
│ ├── meeting-notes/ ← skill-level logs from write-logs
│ │ └── skill.log
│ └── summarize-folder/
│ └── skill.log
├── Agent/
│ └── folder-scout/
│ └── agent.log
├── Bash/ ← tool-level logs from hooks (no sub-name)
│ └── hook.log
└── session/
└── hook.log
The type field from the caller determines the parent folder (Skill, Agent, or Command).
The component field determines the sub-folder name.
$ARGUMENTS — a structured block containing the entries to logThe caller will provide a block like:
component: meeting-notes
type: skill
entries:
- level: DEBUG
message: "[RAW INPUT] the raw input text..."
- level: INFO
flag: --decision
message: "Considered: X. Chose: Y. Why: Z"
- level: DEBUG
message: "[OUTPUT] the formatted output..."
stdio:
- tool: Bash
stdin: "the command that was run"
stdout: "what the command printed to stdout"
stderr: "what the command printed to stderr"
- tool: Read
stdin: "file path requested"
stdout: "file contents returned"
type value | Parent folder |
|---|---|
| skill | Skill |
| agent | Agent |
| command | Command |
Parse the component name, type, entries, and stdio from $ARGUMENTS.
Determine the log directory using the type-to-folder mapping:
type: skill + component: meeting-notes → logs/Skill/meeting-notes/type: agent + component: folder-scout → logs/Agent/folder-scout/type: command + component: deploy → logs/Command/deploy/For each entry, write a log line:
DIR=logs/<ParentFolder>/<component>
mkdir -p "$DIR"
echo '{"timestamp":"<ISO-8601>","component":"<component>","type":"<type>","level":"<level>","event":"<log|decision>","message":"<message>"}' >> "$DIR/<type>.log"
echo '{"timestamp":"<ISO-8601>","component":"<component>","type":"<type>","level":"DEBUG","event":"stdio","tool":"<tool>","stdin":"<stdin>","stdout":"<stdout>","stderr":"<stderr>"}' >> "$DIR/<type>.log"
Truncate any stdin/stdout/stderr value longer than 2000 characters to the first 2000 characters followed by ... [truncated]. This prevents log files from growing unbounded.
Do not produce any terminal output. Redirect all command output to /dev/null except the log file writes themselves.
Return a one-line confirmation: Logged <N> entries and <M> stdio captures to logs/<ParentFolder>/<component>/
stdio section, skip stdio logging — it is optional.npx claudepluginhub getbonzo/agentic-jumpstart --plugin agentic-jumpstartProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.