apify-collect

Apify collect

Gather fresh job URLs into ~/.dear-hiring-manager/urls.txt (the single input to /dear-hiring-manager:batch) by running an Apify job scraper — purpose-built, far fresher and more complete than web search.

Config

• APIFY_TOKEN — your Apify API token. Read from ~/.dear-hiring-manager/.env (APIFY_TOKEN=...) or the APIFY_TOKEN env var. Never put it in the repo, in a URL, in command output, or in any log (.env is gitignored).
• APIFY_ACTOR — the Apify job-scraper actor to run (a LinkedIn / Indeed / ATS-board actor of your choice). Set it in the same .env, or ask the user which actor to use.
• APIFY_INPUT (optional) — a JSON template for that actor's input, using its own field names (actor input schemas differ — there is no universal search/location contract). Put it in ~/.dear-hiring-manager/apify.json (or point to a file). If it's absent, fetch the actor's input schema first (GET https://api.apify.com/v2/acts/$APIFY_ACTOR with the Bearer header) and build the input from the schema + the profile targets, then confirm with the user before running.

Procedure

1. Load the token without echoing it, e.g. APIFY_TOKEN=$(grep -m1 '^APIFY_TOKEN=' ~/.dear-hiring-manager/.env | cut -d= -f2-) (fall back to the env var); same for APIFY_ACTOR. If the token is missing, stop and tell the user to put it in ~/.dear-hiring-manager/.env. Never print the token or a command with it expanded.
2. Build the search from the profile: Desired job title(s) + Target experience level + preferred locations / remote. Only ask the user if those profile fields are blank.
3. Build the actor input from APIFY_INPUT (the actor's own field names) or, if unset, from the fetched input schema + profile targets (step above). Run the actor synchronously with header auth, never the URL:

   curl -sS -X POST \
     -H "Authorization: Bearer $APIFY_TOKEN" \
     -H "Content-Type: application/json" \
     -d "$APIFY_INPUT" \
     "https://api.apify.com/v2/acts/$APIFY_ACTOR/run-sync-get-dataset-items"
   

   No ?token= in the URL; do not echo the token or the expanded command.
4. Extract the application URL per returned item. Do NOT assume a field name — actor outputs differ. Scan each item for URL-valued fields and pick the direct company/ATS apply link (a field like applyUrl / externalApplyUrl / companyApplyUrl, or any Greenhouse/Lever/Ashby/Workday URL), preferring it over the aggregator jobUrl (linkedin.com / indeed.com). Skip items with only an Easy-Apply / no external URL. If no item yields a usable URL, tell the user the actor's output shape doesn't expose one.