Job Tracker

Job Tracker

Keeps two plain-text project files in sync with the user's real-world job search:

• List_of_targets -- one company name per line. The watch list.
• Applied_list -- one line per application, format Company - Job Title or Company - Job Title (rejected) or Company - Job Title (interview). No tag means "applied, no outcome yet."

All the actual matching/formatting logic lives in scripts/update_tracker.py so results are deterministic -- don't hand-edit the files yourself, always go through the script.

Step 1: Work out what the user is telling you

From the user's message, extract:

If the user's message is genuinely ambiguous about which company or role they mean (e.g., they've applied to the same company for multiple roles and it's unclear which one just got the update), ask a single clarifying question before running the script rather than guessing.

Step 2: Locate the source files

These are normally Claude Project knowledge files mounted read-only, e.g. /mnt/project/List_of_targets and /mnt/project/Applied_list. Locate them (check the project files list / uploaded files for the current conversation). If you can't find them, ask the user to point you to them or paste their current contents.

Step 3: Run the script

python3 scripts/update_tracker.py \
  --targets "/mnt/project/List_of_targets" \
  --applied "/mnt/project/Applied_list" \
  --out-dir "/home/claude/tracker_update" \
  --company "Skydio" \
  --title "Autonomy Engineer" \
  --status interview

• Company-only mode: omit --title entirely (and therefore --status). The script will only touch List_of_targets and leave Applied_list completely alone.
• Applied with no outcome yet: give --company and --title, omit --status.
• Rejected / interview: give all three flags.
• The script never edits the input files in place -- it always writes fresh copies to --out-dir. This is safe to run even though the source files under /mnt/project/ are typically read-only.
• The script already handles: not creating a duplicate target-list entry, not creating a duplicate Applied_list line for the same company + title (it updates the status on the existing line instead), and matching "Company (suffix)" style target entries like "Dracoe (HQ)".

Step 4: Report back and hand off the files

1. Read the script's printed summary and relay it to the user in plain language -- what got added, what got updated, what was already there.
2. Use your file-presentation tool to surface the two updated files from --out-dir.
3. Always remind the user of this: Project knowledge files are read-only copies. These updated files need to replace the old ones in their Project (delete the old List_of_targets / Applied_list from Project knowledge, upload the new ones) for the change to actually persist -- running this skill does not silently update the Project on its own.

Edge cases

• Company not found, only a company name given: append it to List_of_targets, say so plainly, don't touch Applied_list.
• Company + title already exists in Applied_list with a different or no status: the script updates that line's status in place rather than adding a second line for the same role. Point this out in your summary (e.g. "moved Skydio - Autonomy Engineer from no tag to (interview)").
• User gives a status word that isn't "rejected" or "interview" (e.g. "ghosted", "withdrew", "offer"): ask them whether it should map to one of the two supported tags, or be left untagged, rather than inventing a new tag style the rest of the file doesn't use.
• Multiple roles at the same company in one message: run the script once per role.