Reserving analysis
Skill cas-team-analyst/team-analyst/skills/reserving-analysis
Actuarial reserving analysis workflow. Use when the user requests reserving. Do NOT use for peer review (separate skill exists).From its SKILL.md
npx -y skills add cas-team-analyst/team-analyst --skill reserving-analysisAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 9 stars9 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.4 KB, as published. Nobody here has run it
Reserving Analysis Workflow
Startup Steps
Follow Every Time BEFORE reading PROGRESS.md
- Confirm to the user which version of the skill this is: v20260601-04
- Determine whether you're running as a Cowork agent (VM/filesystem access to the user's machine) or a Claude Chat agent (sandboxed workspace only, no access to the user's local folders — e.g. running as a skill uploaded directly into claude.ai or the desktop Chat app).
- Cowork: Identify and mount the project folder. Do not use your own folder (may cause max path length issues), ask the user to select one. Obtain permission to modify it. Identify and mount the plugin skill folder (if applicable, NOT the anthropic-skills folder) so you can copy files from it into the project folder.
- Chat: User folders aren't accessible or relevant. Set up your own workspace folder for this analysis instead of asking the user to select one. This skill's
assets/,scripts/, andagents/are already available to you directly — there is no separate plugin skill folder to locate or mount. Ask the user to upload the input data file(s) into the chat (at least one loss/claim triangle; optionally exposures, prior selections, expected loss rate/frequency) instead of asking them to point you to a folder.
- Check for existing
PROGRESS.mdfile in the project folder and use it to understand what the next step is. - Only if no
PROGRESS.mdexists:- Copy PROGRESS.md, REPORT.md, and REPLICATE.md from
assets/to project directory. Use copy commands, do NOT read and write the files. - Start steps in PROGRESS.md. Mark them in progress/complete as you go.
- Copy PROGRESS.md, REPORT.md, and REPLICATE.md from
Then, process to complete each step:
- Mark status as "In Progress [yyyy-mm-dd]"
- Inform the user about the step.
- Perform the requested action. Keep the user informed as you work.
- When complete
- Update REPORT.md
- Update REPLICATE.md
- Mark the step as complete with [X]
- Move on to the next step.
Cowork Agent Guidelines
Applies only when running as a Cowork agent (see Startup Steps).
- File operations: To work from assets file: mount skill folder first, then copy files with
cp {BASE_DIRECTORY}/{skill_path} {PROJECT_FOLDER}/{target_path}. DO NOT usecreate_fileor Write tool for copying files from assets. Convert Windows paths to Unix:C:\Users\...→/mnt/c/Users/...If copy fails, STOP and debug. - Fix bugs with targeted edits AFTER copying - never rewrite entire files.
- Cache out of date? Suggest close/reopen Cowork.
- Never use unicode symbols in commands.
Claude Chat Agent Guidelines
Applies only when running as a Claude Chat agent (see Startup Steps).
- No user folders to mount and no plugin skill folder to locate — this skill's files are already part of your context. Copy
assets/templates into your own workspace folder normally. - Get input data via chat upload, not a file path.
Supporting Files
Skill folder structure:
assets/- Templates (PROGRESS.md, REPORT.md, REPLICATE.md, welcome message) and selection framework guidescripts/- Numbered Python workflow scripts (1a-7) and modules/ subdirectory with shared utilitiesagents/- Selector subagents for LDF, tail curve method, and ultimate selections (rules-based and open-ended variants)